The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo Nga Thi Viet Nguyen and Felipe F. Dizon November 2017 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo Nga Thi Viet Nguyen and Felipe F. Dizon Photo Credits Cover page (top): © Georges Tadonki Cover page (center): © Curt Carnemark/World Bank Cover page (bottom): © Curt Carnemark/World Bank Page 1: © Adrian Turner/Flickr Page 7: © Arne Hoel/World Bank Page 15: © Adrian Turner/Flickr Page 32: © Dominic Chavez/World Bank Page 48: © Arne Hoel/World Bank Page 56: © Ami Vitale/World Bank Acknowledgments This study was prepared by Nga Thi Viet Nguyen The team greatly benefited from the valuable and Felipe F. Dizon. Additional contributions were support and feedback of Félicien Accrombessy, made by Brian Blankespoor, Michael Norton, and Prosper R. Backiny-Yetna, Roy Katayama, Rose Irvin Rojas. Marina Tolchinsky provided valuable Mungai, and Kané Youssouf. The team also thanks research assistance. Administrative support by Erick Herman Abiassi, Kathleen Beegle, Benjamin Siele Shifferaw Ketema is gratefully acknowledged. Billard, Luc Christiaensen, Quy-Toan Do, Kristen Himelein, Johannes Hoogeveen, Aparajita Goyal, Overall guidance for this report was received from Jacques Morisset, Elisée Ouedraogo, and Ashesh Andrew L. Dabalen. Prasann for their discussion and comments. Joanne Gaskell, Ayah Mahgoub, and Aly Sanoh pro- vided detailed and careful peer review comments. Acknowledgments iii Abbreviations and Acronyms AIDS Acquired Immune Deficiency Syndrome CGIAR Consultative Group for International Agricultural Research CMU Country Management Unit ECOWAS Economic Community of West African States EMC Continuous Multi-Sectoral Survey (Enquête Multisectorielle Continue) EMICOV Integrated Modular Household Well-Being Survey (Enquête Modulaire Intégrée sur les Conditions de Vie des Ménages) ENV Household Living Standards Survey (Enquête sur le Niveau de Vie des Ménages) FAO Food and Agriculture Organization PFR Rural Land Plans (Plans Fonciers Ruraux) FEWSNET Famine Early Warning Systems Network GADM Global Administrative Area Database GDP Gross Domestic Product HI Herfindahl Index HIV Human Immunodeficiency Virus ICT Information and Communication Technology IFPRI International Food Policy Research Institute LMI Low and Middle Income NEG New Economic Geography OECD Organisation for Economic Co-operation and Development PAD Project Appraisal Document PFR Rural Land Use Plan (Plan Foncier Rural) PID Project Information Document PPP Purchasing Power Parity QUIBB Basic Well-Being Indicator Questionnaire (Questionnaire des Indicateurs de Base du Bien-être) R&D Research and Development SCD Systematic Country Diagnostic SHIP Survey-Based Harmonized Indicators Program SMS Short Message Service SSA Sub-Saharan Africa TFP Total Factor Productivity UN United Nations WAEMU West African Economic and Monetary Union WDI World Development Indicator Abbreviations and Acronyms v Table of Contents Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii Abbreviations and Acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v Executive Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi Chapter 1: Location and Prosperity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 Motivation and Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Regional Context . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Chapter 2: Geography of Welfare—Three Building Blocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 Natural Endowment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 Agglomeration Economies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 Market Access . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 Chapter 3: Spatial Disparities in Welfare and Poverty . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 Leading and Lagging Areas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 Poverty Rates, Poverty Mass, and Poverty Density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 Access to Services . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 Quality of Life and Characteristics of Poor People . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 Chapter 4: Geographical Differences in Agricultural Activity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 Employment in Agriculture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 Agricultural Productivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 Assets, Inputs, and Output Markets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 Chapter 5: Putting It All Together . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 Relationship between Welfare and Agricultural Productivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 Three Sets of Explanatory Variables for Three Building Blocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 Correlates of Welfare . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 Correlates of Agricultural Productivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 Chapter 6: Policy Recommendations and Further Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 Urbanization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 Agricultural Productivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58 Fiscal Transfers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 Safety Net Programs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 Limitations and Further Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 Table of Contents vii Appendix A: Agroecological Zones . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68 Appendix B: Market Accessibility Index—Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 Appendix C: Extra Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73 Appendix D: Summary of Findings on Agricultural Activities across Countries, by Zone . . . . . . . . 83 Appendix E: Agricultural Data—Notes on Model Construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85 Maps, Tables, Figures, and Boxes Map 1.1: Climate Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 Map 2.1: Abundant in Terms of Precipitation Levels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 Map 2.2: Agroecological Zones, by Country . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 Map 2.3: Concentration of Agglomeration Economies in the South . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 Map 2.4: Road Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 Map 2.5: Concentration of High Market Access in the South, around Capitals and Economic Capitals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 Map 3.1: Cluster of Leading Areas in the South, around the Capitals and the Economic Capitals, or along Country Borders . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 Map 3.2: In Three out of Four Countries, the North is Remarkably Poorer than the South . . . . . . . . . 21 Map 3.3: Lower Poverty Rates in Areas around the Capitals and the Economic Capitals, and along the Country Border . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 Map 3.4: High Poverty Density in and around the Capitals and the Economic Capitals, and in the South . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 Map 3.5: High Variation in Access to Public Services across Space (except cellphone coverage) . . . 25 Map 3.6: Low Service Coverage in Areas with High Poverty Incidence . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 Map 3.7: Higher Diversity in Food Consumption Basket and Lower Food Share from Own Production in the South . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 Map 3.8: Variation in Key Food Consumed across Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 Map 4.1: Employment in Agriculture Relative to Other Sectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 Map 4.2: Cash Crops across Agroecological Zones . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 Map 4.3: Cash Crops across Côte d’Ivoire’s Agroecological Zones . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 Map 4.4: Maize Yields across Agroecological Zones . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 Map 4.5: Cash Crop Yields across Agroecological Zones . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 Map 4.6: Use of Inputs and Farm Land across Agroecological Zones . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 Map 4.7: Land Tenure Security across Agroecological Zones . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 Map 4.8: Sale of Agricultural Produce . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 Table 1.1: Economic per Capita Growth by International Standards . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Table 1.2: Geographical Data Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 Table 3.1: Better Housing Conditions for Poor Households in Urban Areas or in Favorable Agroecological Zones . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 Table 3.2: Fewer Family Members, Lower Dependency Rates, More Likely to be Female-Headed Households, and Less Likely to Have No Education among Poor Households in Urban Areas and Favorable Zones . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 Table 5.1: Statistical Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 viii The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo Table 5.2: Factors Associated with Spatial Differences in Poverty: Coastal Location, Population Density, and Market Access . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 Table 5.3: Role of Geographical Differences in Agricultural Productivity, Natural Endowments (temperature, latitude, elevation, coastal location), and Spending on Fertilizer . . . . . . . . . . . . . . . . . . 54 Figure 2.1: Higher Share of Population Living in Urban Areas but Slower Urban Population Growth by African Standards . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 Figure 3.1: Large Wealth Gap between Leading and Lagging Areas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 Figure 3.2: Many Leading Areas in Low-density Locations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 Figure 3.3: Low Market Access in Many Leading Areas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 Figure 3.4: Majority of the Poor Living in Low-density Areas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 Figure 3.5: Large Gaps in Public Service Coverage between the Most Sparsely and Most Densely Populated Areas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 Figure 4.1: Percentage of Population Engaged in Agriculture, by Poor and Nonpoor . . . . . . . . . . . . . . . 33 Figure 4.2: Crops Grown by Poor and Nonpoor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 Figure 5.1: Correlation between Poverty and Agricultural Productivity . . . . . . . . . . . . . . . . . . . . . . . . . . 50 Box 4.1: Summary of the agriculture sector based on various World Bank documents (Project Information Document [PID], Project Appraisal Document [PAD], and Systematic Country Diagnostic [SCD]) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 Box 4.2: Background on land reform in Benin, Côte d’Ivoire, and Burkina Faso . . . . . . . . . . . . . . . . . . . 45 Table of Contents  ix Executive Summary West Africa is at the heart of Africa’s transforma- cornerstone of the economy in this subregion, tion. With a gross domestic product (GDP) growth the report explores geographical differences rate of more than 5 percent annually, it is the fast- in agricultural activity; est growing region in the continent. Yet poverty 3. It quantifies the roles of natural endowment, rates remain high, even by African standards. In agglomeration economies, and market access the four West African countries covered in this in determining the spatial distribution of wel- report, namely Benin, Burkina Faso, Côte d’Ivoire, fare and agricultural productivity; and and Togo, nearly half of the population lives on less than US$1.90 a day at 2011 purchasing power 4. It suggests a number of policy guidelines that parity (PPP). This means over 25 million people live may help improve shared prosperity across in extreme poverty. space. How is it possible that the subregion’s high eco- However, we acknowledge that since poverty nomic growth cannot translate into higher levels is a multidimensional concept, many other fac- of prosperity? The answer lies in where one looks, tors could potentially contribute to the observed as national averages often mask large disparities within-country inequality, yet they are not cov- at subnational levels. ered by the scope of this study. These may include social elements such as ethnicity, nutrition, and Recent literature on the new economic geography health; economic conditions such as prices and suggests that within-country disparities may be a markets; and political dimensions such as institu- natural outcome of the development process. As tions and conflict. a country develops, economic activity clusters in regions endowed with more favorable agroecologi- cal conditions, more abundant natural resources deposits, or simply a better location. More eco- A Tale of Two Regions nomic opportunities in turn attract more people in search of jobs, which consequently increases pop- In terms of agroecological endowment, two dis- ulation density in one area over another. Arguably, tinct groups emerge within the subregion. Being the higher concentration of people and economic the most northerly and the only landlocked coun- activity leads to economies of scale. Such ben- try in the set, Burkina Faso has a noticeably differ- efits can be further enhanced with the existence ent agroecosystem, being generally drier and less of market access for products, labor, and ideas, fertile. It also has the most dispersed population which continues to boost these regions’ income in the subregion. The rest—Benin, Côte d’Ivoire, and attractiveness to people and firms. This vir- and Togo—are coastal and located at the same tuous cycle of development makes it difficult for latitude and therefore share similar agroecological poor regions to catch up. endowments. Of the four countries, Togo has the highest population density. This report aims to assess the spatial disparities in economic development along four important Within each of the three coastal countries, there dimensions: also exists a tale of two regions, with the South having more favorable conditions for agricultural 1. It provides stylized facts of the underly- activities, access to the sea, and higher popula- ing forces behind within-country inequality, tion density than the North. Thus, the pattern of namely natural endowment, agglomeration market access is similar to that for natural endow- economies, and market access. These are the ment and agglomeration, with higher levels of three building blocks of the economic geogra- market access concentrating in the South. phy literature; The picture is slightly different in landlocked 2. It examines spatial disparities in welfare Burkina Faso. While it shares a similar geographi- and poverty. As the agricultural sector is a cal pattern of agroecological characteristics as Executive Summary xi the one found in the neighboring countries (i.e., typical household in a leading area may consume North vs. South), its population and market as much as seven times more overall than a similar access concentrate only in the Central region, household in a lagging area. This gap is highest in home of the two largest cities: Ouagadougou and Benin and lowest in Burkina Faso. Bobo-Dioulasso. Notably, many of the leading areas have not yet maximized the benefits of agglomeration econ- omies. This observation is especially clear in A Tale of Two Economies Burkina Faso and Côte d’Ivoire, where approxi- mately half of the leading regions are located in In the three coastal countries, the North is mark- low-density areas. This suggests scope for greater edly poorer and has a larger share of population concentration of economic activities and labor in employed in the agriculture sector than the South. these locations in order to further take advan- In Togo particularly, poverty in the far North may tage of economies of scale and boost economic be more than three times as high as in the far development. South. However, the pattern of poverty is reversed Moreover, in each country, there exists geographi- in Burkina Faso, in part because of the possession cal pockets of poverty that may be resistant to of livestock among northern residents. policy-induced changes. These lagging areas are The spatial distribution of crops grown also var- characterized by a combination of high poverty ies between the North and South. Cash crops, or rates and a low number of poor people per square crops produced for commercial value, are more kilometer. As a result, the unit cost of a poverty prevalent in the South (except cotton) in terms of targeted program may be extremely high in these the proportion of farmers growing them. However, areas. Given budget constraints, the government the geographical coverage of cotton production may not be able to reach this population group. is quite different and resembles a belt cover- ing the southern parts of Burkina Faso and the northern parts of Benin, Côte d’Ivoire, and Togo. Interestingly, this cotton belt overlaps with areas Spatial Disparities of higher agglomeration distant from the capital Explained city of each country. There has been a long ongoing debate in the eco- Looking beyond monetary poverty and agricultural nomic geography literature on whether a location’s activity, the quality of life of the poor as measured levels of per capita income and other economic by the extent of food intake diversification, access dimensions are determined by geographical and to basic services, and housing conditions increases ecological variables. Many researchers have pro- significantly from North to South. In other words, vided evidence supporting the view that such links given two poor individuals with similar incomes, are strong, while others have argued that the role the one living in the South enjoys a more diverse of geography in explaining spatial patterns of per food basket, has a higher chance of having access capita income operates through various direct to electricity and sanitation, and has a higher channels (e.g., productivity and trade) or indirect probability of living in a house with either a con- channels (e.g., choice of political and economic crete roof or brick walls than his fellow citizen in institutions), with little direct effect of geography the North. These patterns are consistent across on incomes. all four countries. How therefore does this play out in the case of At subnational levels, differences in welfare are Benin, Burkina Faso, Côte d’Ivoire, and Togo? even more pronounced between leading and lag- ging locations.1 A location is defined as “leading” if As it turns out, except for being coastal or land- per capita consumption for an average household locked, agroecological characteristics do not living there is higher than the national average. A appear to be directly associated with a location’s per capita income. The relationship between geog- raphy and welfare is indeed being mediated by agglomeration economies and market access. 1 Subnational levels consist of communes in Benin, provinces in Burkina Faso, departments in Côte d’Ivoire, and prefectures in Togo. In other words, locations favorable to growth xii The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo will either attract people or experience stronger growth and political stability. How can our find- population growth and at the same time receive ings help policy makers reduce geographical dif- investments in infrastructure. Thus, when con- ferences in welfare while boosting growth? Based trolling for population density and market access, on our analysis, we propose four broad policy the correlation between welfare and geographical recommendations: variables (except for being located along the coast) 1. Urbanization: We find that many of the lead- is no longer significant. ing areas have not yet maximized the ben- If natural endowment plays any role in explain- efits of agglomeration economies, especially ing spatial disparities in welfare, the key factor is in Burkina Faso and Côte d’Ivoire. Based on location near the coast. Given two locations with the new economic geography literature, there the exact same population density and market is scope for increasing concentration in eco- access, the one on the coast is 21 percent richer nomic activities and labor in these areas to than the one located inland. The fact that the eco- further take advantage of economies of scale nomic benefits of coastlines remain strong (i.e., and boost economic development. However, they have not been arbitraged away by migration it is important to consider complementary or increased market access) reveals the untapped policies to urbanization, including removing potential for economic development provided by barriers to labor mobility so that people can access to international trade for the three coastal migrate to leading areas where demand for countries (Benin, Côte d’Ivoire, and Togo). labor and productivity are higher, and invest- ing in urban infrastructure and the provision However, the story is quite different when look- of public services to accommodate a poten- ing at agricultural productivity measured by tially larger number of users. maize yields. This report focuses on maize yields because this food crop is fairly prevalent across 2. Increasing agricultural productivity: Not all all four countries and is found in most areas in rural families can move to urban locations. each country. This allows for some degree of For those staying in the agriculture sector comparability for yields across zones and across in rural areas, policy makers should consider countries. improving welfare by increasing agricultural productivity. Potential areas of improvement What is notable is the persistence of the correla- include land tenure, irrigation, use of farm tion between geography and agricultural produc- inputs such as fertilizer, and research and tivity regardless of whether population density, development. market access, or farm inputs are taken into account. In contrast to the new economic geog- 3. Fiscal budget transfers: Geographical pock- raphy literature suggesting that agglomeration ets of poverty exist where the costs of reach- economies and market access can help farmers ing the poor are very high. These areas are take advantage of better prices, a wider selection characterized by a combination of high pov- of agricultural inputs, and better markets for har- erty rates and low poverty density. Another vested crops, this link is weak in the subregion. This set of lagging areas with little prospect of finding implies that there may in fact be two types growth consists of those with unfavorable of agriculture: a subsistence agriculture, whereby agroecological characteristics and limited most crops are cultivated for home consumption opportunities to diversify into nonagricul- and where investments are less sensitive to mar- ture sectors. Our quantitative analyses ket access, and a commercial agriculture, which show a persistent link between agroecologi- might concentrate along coastlines and benefit cal endowment and agricultural productivity from higher investment in inputs. regardless of whether agglomeration, market access, or farm inputs are taken into account. Our findings imply that some lagging areas may not be able to improve their welfare after What Can Be Done? all. This may call for pro-poor fiscal transfers through a system of inter-region transfers Within-country disparities can be a potential to ensure equity across leading and lagging source of tensions between lagging and leading areas. locations and may affect the country’s overall Executive Summary   xiii 4. Safety net programs: Not all poor people, low-density areas in a cost-effective way. especially the vulnerable, can benefit from Moreover, safety net programs should be part the policies proposed above. Thus, the need to of an overarching poverty reduction strategy maintain strong safety net programs target- consisting of interacting with and working ing the poor and vulnerable remains strong. alongside urban policy, agricultural produc- New technologies such as e-vouchers and tivity boosting programs, and other policies mobile transfers make it possible for such aimed at eradicating poverty and reducing programs to reach targeted beneficiaries in vulnerability. xiv The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo Chapter 1 Location and Prosperity Motivation and Objectives This study covers four countries in West Africa under the Country Management Unit Location is the most critical predictor of a per- (CMU) AFCF2, namely Benin, Burkina Faso, son’s welfare (World Bank, 2009). As of today, a Côte d’Ivoire, and Togo, and makes use of four child born in Togo is expected to live nearly 20 years recently collected household consumption less than a child born in the United States. More- surveys. Given the data limitations, we focus over, the child will earn a tiny ­ fraction—less than on a static analysis of economic geography in 3 percent—of what her American counterpart will this subregion. While our discussion is centered earn (World Bank, 2017). around within-country inequalities as these are more relevant to each respective government, Such disparities in income and living standards we touch upon some aspects of cross-country within a country are just as unsettling. An urban differences in order to provide a regional con- inhabitant in Togo’s capital, Lomé, has a 16 percent text. We also emphasize that our analysis focuses chance of being poor, and a 90 percent chance mainly on the spatial distribution of welfare and of having access to electricity. However, these poverty. probabilities are reversed for a person from a rural district in far northern Oti prefecture, where resi- To understand the driving forces behind differ- dents have an 80 percent chance of falling into ences of welfare and poverty across space, we poverty and a mere 13 percent chance of having base our analysis on the NEG literature and high- access to electricity. light its three building blocks: natural endowment, agglomeration economies, and market access. In As pointed out in the World Development addition, we explore the spatial distribution of sev- Report—Reshaping Economic Geography (World eral key elements that intertwine with poverty, Bank, 2009), such within-country disparities can including economic activity, agricultural produc- pose a major challenge for policy makers as they tivity, household demographics, and access to ser- present a potential source of increasing tensions vices (Banerjee and Duflo, 2007). We acknowledge between poorer and richer areas. Moreover, if that since poverty is a multidimensional concept, these spatial inequalities persist or widen, they many other factors could potentially contribute can potentially affect a country’s future growth to observed within-country inequality, yet are not and political stability. covered by the scope of this study. These may Spatial differences in economic development include social elements such as ethnicity, nutri- have long been the subject of study, with a tion, and health; economic conditions such as history dating back to the 4th century b.c. and prices and markets; and political dimension such expanding after World War II due to uneven post- as institutions and conflict. war economic recovery and development. Until Our objectives are: the 1980s, the study of economic geography was under scrutiny because it undermines the notion 1. To provide stylized facts relevant to the three of equal opportunity among individuals. How- building blocks of the economic geography ever, over the past decades, the field has regained literature: natural endowment, agglomeration attention in mainstream development debates economies, and market access; thanks to new theories of economic growth and 2. To examine static spatial disparities in welfare empirical research in this field (Hausmann, 2001). and poverty together with relevant develop- The recent literature on the new economic geog- ment indicators such as poor household raphy (NEG) implies that within-country dispari- demographics, access to services, economic ties may be a natural outcome of the development activity, and agricultural productivity; process, and once established, can be persistent 3. To quantify the roles of natural endowment, and insensitive to policy-induced changes (see market access, and agglomeration economies Fujita, Krugman, and Venables, 1999; Puga, 1999; in determining the spatial distribution of wel- Fujita and Thisse, 2002; World Bank, 2009). Thus, fare; and from a policy maker’s perspective, the success of any government policy aimed at improving shared 4. To suggest a number of policy guidelines that prosperity across locations crucially depends on may help improve shared prosperity across what drives the observed spatial inequality. space. 2 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo The report is organized as follows. The rest of the population have no access to improved sanita- Chapter 1 provides a glance at the regional con- tion (World Bank, 2017). Along key factors such as text (i.e., where these countries stand in the global institutional quality, labor productivity, and human economy) as well as an overview of the data used. capital, the region’s geographical characteristics Chapter 2 introduces the three building blocks are often considered to be a key constraint on its of the economic geography literature: natu- economic development. ral endowment, agglomeration economies, and The four West African countries covered in this market access. Chapter 3 presents stylized facts report (Benin, Burkina Faso, Côte d’Ivoire, and about the subregion’s spatial disparities, focus- Togo) lie mostly in the tropical savannah climate ing on welfare, poverty, access to services, and area (Map 1.1), a common geographic disadvan- profiles of the poor. Chapter 4 provides details tage identified among countries lagging behind in of the geographical distribution of agricultural economic development. Hausmann (2001) shows activity given the important role it plays in the that, on average, annual economic growth rates subregion. Chapter 5 uses the NEG framework in tropical nations are between one-half and a full to explore correlates of the observed inequalities percentage point lower than in temperate coun- across space. Finally, Chapter 6 concludes with a tries. In addition, countries located in tropical policy discussion. areas often show more skewed income distribu- tion and poorer health conditions than their non- tropical counterparts. Regional Context As shown in Table 1.1, these four West African countries are poor by international standards. The majority of the world’s extreme poor, those Not only is their per capita GDP lower than the living on less than US$1.90 a day at 2011 pur- average of Low and Middle Income (LMI) countries, chasing power parity (PPP), are concentrated it is also lower than African averages. Even though in Sub-Saharan Africa (SSA). Within the conti- these economies, especially for Burkina Faso and nent, West Africa is home to some of the poorest Côte d’Ivoire, have grown at the impressive rate of nations, where approximately half of the popula- approximately 5 percent per year, annual growth tion lives in poverty, and over three quarters of in per capita GDP still falls behind those of their Map 1.1  Climate Classification Source: Kottek et al., 2006. Location and Prosperity 3 Table 1.1  Economic per Capita Growth by International Standards Levels GDP per GDP, 2015 Capita, GDP Density Population $1.90 (PPP, 2015 2015 ($/km2, Population Density Poverty $billion) (PPP $) thousands) (million) (ppl/km2) Rate (%) Benin 22 2,057 198 11 96 67.8 Burkina Faso 31 1,696 112 18 66 43.7 Côte d’Ivoire 80 3,514 251 23 71 27.9 Togo 11 1,460 196 7 134 49.2 Sub-Saharan 3,718 3,714 157 1,001 42 41.0 Africa Low & Middle 61,047 9,911 645 6,159 65 12.6 Income Annual Growth (2010–2015) (percent) (percent) (percent) (percentage point) Benin 4.3 1.5 2.7 3.67 Burkina Faso 5.9 2.8 3.0 –2.32 Côte d’Ivoire 5.8 3.3 2.4 –0.16 Togo 4.8 2.0 2.7 –1.26 Sub-Saharan 4.3 1.5 2.8 –1.56 Africa Low & Middle 5.2 3.9 1.3 –1.93 Income Source: World Bank, 2017. peers, in part because of relatively fast population above US$3,500 PPP per year, one in every four growth. persons still lives on less than US$700 annually. A possible explanation for this mismatch between A related fact is the high levels of population population density and economic prosperity may density across all four countries. Within this be the vast unevenness in economic development subregion, Togo is the densest, with the number within a country’s border. Chapter 3 will explore of people per square kilometer being about three this aspect further. times higher than SSA averages, and twice the average of LMI countries. Population density levels in the other three countries are also well above international averages. Data Interestingly, the subregion’s relatively high As discussed above, national level comparisons population density and high GDP density mask large disparities at subnational levels. This (defined as GDP per square kilometer), which section describes the data used to explore within- are often considered favorable elements for eco- country disparities in welfare, poverty, and other nomic development (World Bank, 2009), do not development indicators. translate into higher levels of prosperity. In three out of four countries—Benin, Burkina Faso, and Togo—about half of the population lives in extreme Statistical Data poverty, a rate that is even higher than African This study makes use of four recently collected averages. Even in Côte d’Ivoire, the only middle- household consumption surveys: the Benin Inte- income country in the group, with a per capita GDP grated Modular Household Well-Being Survey 4 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo (Enquête Modulaire Intégrée sur les Conditions de able, wide variance (e.g., outliers). To arrive at a Vie des Ménages, EMICOV 2015), the Burkina Faso balance between data and administrative coher- Continuous Multi-Sectoral Survey (Enquête Multi- ence such that conclusions can be useful from a sectorielle Continue, EMC 2014), the Côte d’Ivoire policy and development perspective, we focus on Household Living Standards Survey (Enquête sur second subnational level data (i.e., communes for le Niveau de Vie des Ménages, ENV 2015), and the Benin, provinces for Burkina Faso, departments Togo Basic Well-Being Indicator Questionnaire for Côte d’Ivoire, and prefectures for Togo). It is (Questionnaire des Indicateurs de Base du Bien-être, important to note that even at these subnational QUIBB 2015). levels, we can limit but not entirely eliminate the two shortcomings discussed above. While earlier household consumption surveys for each country are also available (i.e., Benin EMICOV We rely on the agriculture and land modules con- 2010, Burkina Faso GHS 2009, Côte d’Ivoire ENV tained in these surveys to capture the geography 2011, and Togo QUIBB 2011), their lack of compa- of agriculture from the perspective of households. rability with more recent surveys limits our capa- Unlike administrative data on agricultural pro- bility to observe changes in geographical patterns duction, our approach is more likely to be biased of welfare and poverty over time.2 Therefore, we toward smallholder agriculture instead of large focus on a static analysis of economic geography commercial farms. While livestock is a major in this subregion. source of income for households in Sahel regions, the data are not available across all four countries We also take advantage of a number of harmo- in the subregion, and this dimension is therefore nized data sets from the Survey-Based Harmo- excluded from our analysis. nized Indicators Program (SHIP) produced by the World Bank, which aim to compile in a consistent format consumption aggregates and other house- Geographic Data hold indicators such as demographics and assets To construct a market access index, we used the from household budget surveys in the SSA sub- road network provided by DeLorme (2015). While region. However, our four surveys were collected an ideal index should capture access to all modes only recently and have not been fully processed in of transportation (e.g., air, coast, rail, etc.), we did SHIP at the time of the writing. not have access to such data at the time of writ- In general, household consumption surveys, ing. Thus, our index is limited to reflecting domes- including those used in this report, are designed tic market access to roads. to produce welfare measures and development For our multivariate regressions in Chapter 5, outcomes at the national level and, in some cases, we constructed six continuous agro­ ecological at the first subnational level (e.g., regions). Dis- variables at the administrative unit level for aggregating the data into lower administrative each country:4 temperature, precipitation, soil unit levels may pose two risks: lack of represen- quality, latitude, elevation, and ruggedness. The tativeness, and imprecise estimates.3 On the one temperature variable is a long-run (1960–1990) hand, households who live in a small geographical annual average taken from Hijmans et al. (2005), area and were interviewed for the surveys may and precipitation is taken from HarvestChoice/­ not represent the wider population. On the other, International Food Policy Research Institute (IFPRI) the limited number of households reporting the and University of Minnesota (2016), which mea- information of interest leads to higher odds of sures annual average over the period 1960–2014. ending up with missing information (e.g., access Soil quality is measured as organic carbon soil to improved toilets), or, when information is avail- content (fine earth fraction) at 60–100cm depth taken from HarvestChoice/IFPRI and University of Minnesota (2016). Ruggedness is based on Nunn 2 Household consumption surveys are considered comparable if all and Puga (2012). Elevation, given in meters, is three of the following criteria are consistent across surveys: (i) the taken from Isciences (2008). sample size is nationally representative; (ii) the data were collected during the same period; and (iii) the surveys rely on the same report- ing instrument and reporting period (Beegle et al., 2016). 3 As a rule of thumb, estimates are considered sufficiently precise if the relative standard error (measured as standard error divided by 4 Communes for Benin, provinces for Burkina Faso, departments the mean) is less than 10 percent. for Côte d’Ivoire, and prefectures for Togo. Location and Prosperity 5 Table 1.2  Geographical Data Sources Data Sources Administrative boundaries National statistical services, Global Administrative Area Database (GADM) Agroecological zones (Benin, Togo) Food and Agriculture Organization (FAO) of the United Nations, Togo’s Ministry of the Environment and Forestry Resources Climate zones Kottek et al. (2006) Elevation Isciences (2008) Livelihood zones (Burkina Faso, Côte Famine Early Warning Systems Network (FEWSNET), d’Ivoire) AGRHYMET Population density at subnational National statistical services levels Precipitation HarvestChoice/IFPRI and University of Minnesota (2016) Road network DeLorme (2015) Ruggedness Nunn and Puga (2012) Soil quality HarvestChoice/IFPRI and University of Minnesota (2016) Temperature Hijmans et al. (2005) Except for HarvestChoice/IFPRI and University of Minnesota (2016) data is 5-arc-minute, or roughly Minnesota (2016), these data consist of a 30-arc- 10 3 10 km. Thus, some smaller administra- second grid solution, equivalent to 1 3 1 km, thus tive units will not have data. For such areas, we making it possible to aggregate the data to the impute a given variable as the average of that vari- administrative unit level. For its part, grid resolu- able across its neighboring administrative units tion for HarvestChoice/IFPRI and University of (Table 1.2). 6 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo chapter 2 Geography of Welfare— Three Building Blocks This chapter defines three spatial scales as step- characteristics for agricultural activities, ping stones for the spatial analysis of welfare and while the South is the most endowed. agricultural activity to follow in Chapters 3 and 3. Similarly, agglomeration economies clus- 4. These scales are based on the three building ter mainly in the South. The exception is blocks found in the economic geography literature. landlocked Burkina Faso, where the densest We start with the building block found in the tra- population is in the central region, home of its ditional economic geography literature: natural capital, Ouagadougou. endowment. A region is expected to be better off 4. Market access follows the same pattern. than others if it is endowed with a more favorable Most land areas in the North have very limited agroecosystem, more natural resources, or simply access to markets, while high levels of market a better location. We then add two main elements access concentrate around the capitals, the from any NEG model as well as from any modern economic capitals, and along the coast in the theory of location:5 agglomeration economies (Mar- South. shall, 1920; Krugman, 1991; Porter, 1998; Hender- son, 2014), and access to markets, that is, markets for goods, labor, and ideas (Smith, 1776; Fujita and Thisse, 2002). Finally, we assess what these three Natural Endowment elements look like in each country: Benin, Burkina Faso, Côte d’Ivoire, and Togo. In this section, we examine each country’s natu- ral endowments (i.e., agroecological endowment), The core idea of NEG is that a location is not an standardize existing classifications of agroeco- isolated geographical area but is affected by its rela- logical zones, and regroup these into four broad tionships or connections with neighboring locations. zones ranging from least favorable (Zone 1) to Agglomeration economies ensure that economic most favorable (Zone 4). This recategorization activity is concentrated in areas that are better allows us to overlay patterns of welfare, poverty, located to benefit from increasing returns to scale. agricultural productivity, and economic activity Access to markets then captures the levels of trans- over agroecological zones in a consistent and sys- portation costs and the degree of labor mobility tematic manner across countries. between locations. High levels of market access (i.e., free movement of goods and people across space) Agriculture plays a key role in the four coun- combined with increasing returns to scale will cre- tries of interest. This sector generates about a ate spatial disparities in economic activities, and third of GDP value each year. In Togo, nearly half therefore poverty. In this study, we do not cover of GDP in 2015 came from agriculture alone. In tangible costs such as road tolls or legal require- addition, the agriculture sector provides jobs to ments for residency or non-tangible costs such as about half of the workforce (World Bank, 2017). discrimination or ethnic or religious differences that In an agricultural economy, agroecological endow- may be associated with market access. ments determine not only what crops are planted or livestock is raised in each location but also what Our main findings are: returns can be obtained for any crop harvested or 1. In all four countries, there seems to exist a livestock herded. These consequently affect the tale of two regions—North vs. South for the region’s agricultural productivity and welfare (see, coastal countries (Benin, Côte d’Ivoire, and for example, Diamond, 1997). Togo), and Center vs. the Rest for landlocked Compared to the rest of SSA, West Africa is Burkina Faso. relatively abundant in precipitation, with most 2. Within a country, the North generally of its land in the savannah and grassland areas, has the least favorable agroecological and it benefits from a tropical climate (Map 2.1, Map 2.2, and Appendix A). Within the subregion, two distinct groups emerge. Being the northern- 5 For a detailed discussion of NEG, see, for example, Fujita, Krug- most and the only landlocked country, Burkina man, and Venables (1999); Fujita and Thisse (2002); Baldwin et al., Faso has noticeably different agroecological char- (2003); Brakman, Garretsen, and Van Marrewijk (2009); Combes, Mayer, and Thisse (2008). For major NEG models, see Krugman acteristics, being generally drier and less fertile. (1991); Krugman and Venables (1995); Venables (1996); and Puga The rest—Côte d’Ivoire, Benin, and Togo—are (1999). 8 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo Map 2.1  Abundant in Terms of Precipitation Levels Source: Funk et al., 2015. coastal and located at the same latitude, thus with Zone 1 in Burkina Faso, Benin, and Togo. sharing similar climates and precipitation levels. The area is characterized by savannah, with a mixture of woodland and grassland. A clear picture immediately comes to the fore- front: In all four countries, Zone 1 is in the north, • Togo: This zone has very similar characteristics while Zone 4 is concentrated in the south. Inter- in terms of savannah landscapes and average estingly, Zones 4 in Benin, Côte d’Ivoire, and Togo amounts of rainfall, albeit slightly lower, are also coastal, which is generally considered an compared to Zone 1 in Côte d’Ivoire. advantage for economic development (Map 2.2). Overviews of each zone are detailed below. • Benin: The climate is Sudano-Sahelian with a unimodal rainfall pattern of 700–1,000 Zone 1 represents the Sahelian and Sudanian mm per year. The area is marked with a vast savannah areas, the driest of all four zones within expanse of arable land in ferrosol soil. a country. There is only one rainy season, which is Zone 2 is also characterized by unimodal rainfall also relatively short. However, precipitation levels patterns, albeit with slightly higher precipitation vary considerably across countries, starting from levels and a longer rainy season than in Zone 1. The the lowest in Burkina Faso to the highest in Côte capital of Burkina Faso, Ouagadougou, is located d’Ivoire. in this zone. • Burkina Faso: Zone 1 is typical Sahelian, with • Burkina Faso: This Sudano-Sahelian zone four months of rainfall per year, accumulating receives about 600–800 mm of rainfall per approximately 400–500 mm. The soil is year. However, it has poor quality soil and sandy and of poor quality. faces serious land erosion problems. • Côte d’Ivoire: While the annual precipitation • Côte d’Ivoire: Vegetation types are the same level of 1,000–1,100 mm for five months of the as in Zone 1, i.e., woodlands, grassland, and year is lower than those in other zones within savannah. However, precipitation levels are Côte d’Ivoire, it is the highest when compared higher, at 1,100–1,300 mm per year. The zone Geography of Welfare—Three Building Blocks 9 Map 2.2  Agroecological Zones, by Country Sources: AGRHYMET, 2016; Dixon and Holt, 2010; FAO, 2001, 2009a, 2009b; Ministère de l’Environnement et des Ressources Forestières, 2003, 2014; Vissoh et al., 2004. is typically characterized by flat terrain and • Benin: Being a transitional zone, it has no clear ferrosol soil. distinction between the two rainy seasons. • Togo: The climate is Sudano-Guinean, with The landscape, however, is similar to Zone 2, which is woody savannah with tropical savannah landscapes. ferruginous soil. • Benin: This zone shares similar characteristics Zone 4 has the most abundant rainfall and the with Togo’s Zone 2. most fertile soil. It houses the capital Lomé in Zone 3 receives more plentiful rainfall than the Togo and the economic capitals in Côte d’Ivoire other two zones. and Benin, Abidjan and Cotonou, respectively. • Burkina Faso : The zone transits into a • Burkina Faso: This zone shares a similar savannah ecosystem and a Sudanian climate. ecosystem and climate with Zone 3. However, Precipitation levels are 800–900 mm per year it receives more rainfall, at about 900– (unimodal rainfall curves) along with good soil 1,100 mm per year. quality. The zone is also endowed with large forests and vast areas of animal reserves. • Côte d’Ivoire: The coastal zone receives up to 1,750 mm of rainfall per year. • Côte d’Ivoire: The zone covers two distinct • Togo: This is a coastal zone with a subequatorial ecosystems: a Guinean type in the mountains, climate. However, precipitation levels are and a Sudanian type in the flatlands. Average lower than in other zones, at about 750–1,000 annual rainfall is 1,250–1,500 mm. mm annually. • Togo: This zone has a Guinean climate and is • Benin: This zone has a subequatorial climate, largely made up of the Togo Mountains, which with two rainy and two dry seasons. The soil can reach nearly 1,000 meters in height at type is mostly ferralitic, including relics of Mount Agou. forest. 10 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo Agglomeration Economies four groups: ultra-remote rural, rural, urban, and ultra-dense urban. Density thresholds used to cat- The second building block—agglomeration egorize these groups are based on the Organisation economies—is a vital engine of innovation and for Economic Co-operation and Development— growth and plays a powerful role in explaining OECD (1994), Uchida and Nelson (2010), and Buys, within-country inequality. As countries develop Chomitz, and Thomas (2005). Specifically: over time, economic activity clusters in certain a. Ultra-remote rural localities are defined as locations to take advantage of both economies of those having fewer than 50 people per square scale and knowledge exchanges (Marshall, 1920; kilometer. Krugman, 1991; Porter, 1998; Henderson, 2014). Greater economic opportunities in turn attract b. Rural areas are locations with population den- people who migrate in search of jobs, which con- sity of between 50 and 150 people per square sequently boosts population density in one area kilometer. over another (World Bank, 2009). c. Urban areas are characterized by having popu- As discussed in Chapter 1, this subregion is rela- lation density of between 150 and 300 people tively densely populated by African standards. In per square kilometer. terms of urbanization, as reported by the various d. Ultra-dense urban localities are areas with national statistical services, the share of popula- more than 300 people per square kilometer. tion residing in urban areas in three ­ countries— As shown in Map 2.3, in three out of four coun- Togo, Benin, and Côte d’Ivoire—is relatively high, at tries, namely Benin, Côte d’Ivoire, and Togo, urban over 40 percent of the total population, a rate that and ultra-dense urban localities are concentrated is higher than the SSA average. However, growth mainly in the South, coinciding with the most rates for urbanization are slower, with Burkina favorable agroecological zone (Zone 4) as well Faso having the lowest share of population living in as coastal areas. Landlocked Burkina Faso is an urban areas, at less than 30 percent. Nevertheless, exception in that the densest part of the country the country is catching up, with impressive urban is located in the central region, around the capital population growth of nearly 6 percent annually Ougadougou. (Figure 2.1). Compared to other countries in the subregion, To go beyond the urban-rural dichotomy seen in Burkina Faso has the most dispersed population. the literature and, more importantly, to arrive at Apart from ultra-dense Kadiogo Province, home to a consistent and systematic classification across the capital Ougadougou, the rest of the country is countries,6 we further distinguish localities into made of ultra-remote rural or rural areas. Based on our classification, the country does not even have 6 Each national statistical agency has a different definition of “urban urban localities. In contrast, Togo’s population is locality.” Figure 2.1  Higher Share of Population Living in Urban Areas but Slower Urban Population Growth by African Standards Urban population (% total) Urban population growth (annual %) BFA BEN SSA CIV TGO TGO BEN SSA CIV BFA 0 20 40 60 0 2 4 6   Source: World Bank, 2017. Geography of Welfare—Three Building Blocks 11 Map 2.3  Concentration of Agglomeration Economies in the South Source: National statistical agencies. the densest, with the only ultra-remote rural loca- commonly used modified model of the “classical” tion being found in the Central region, where the approach.7 Togo Mountains lie. As highlighted in NEG literature, access to markets must be considered beyond a coun- try’s border. This element is especially crucial Market Access for landlocked countries such as Burkina Faso. For Burkina goods to reach new markets and for Based on NEG theory, the benefits of agglomera- Burkina people to receive more products from the tion economies can be further enhanced with outside world, there must be good transport con- the existence of good access to markets for nections to neighboring countries. As shown in products, labor, and ideas (Mayer, 2008). A region Map 2.4 several primary roads connect large cities with better market access will attract more eco- in Burkina Faso to their neighbors’ coastal ports nomic activity and labor, leading to an agglomera- in Benin, Côte d’Ivoire, and Togo. tion advantage over time. With increasing returns Among the four countries, landlocked Burkina Faso to scale, the region can then afford to reinvest in has a relatively extensive primary and secondary market access and further reinforce its advantage. road network that extends to all four agroecological This can start a virtuous cycle of development, zones. In contrast, coastal countries such as Côte which is good for economic growth and poverty d’Ivoire and Benin concentrate domestic transpor- reduction but also makes it difficult for disadvan- tation systems in coastal and dense areas (Zone 4) taged regions to catch up (World Bank, 2009). and neglect remote regions (Zones 1 and 2). In this section, we start with a brief glance at the current road network, a core factor in mar- ket access. We then proceed to calculate each 7 For more details of the classical and modified models, see, for country’s market access index based on the most example, Deichmann (1997) and Lall, Shalizi, and Deichmann (2004). 12 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo Map 2.4  Road Networks Source: DeLorme, 2015. To measure market access, we first follow the − t ij− b ( ) classical model in the literature, as follows: MAi = ∑ Pj ⋅ e 2a2 Domestic market access for a given location along � j a road network is a function of the weighted sum where Pj is the population in location j, t ij is travel of populated locations of all other locations dis- time between locations i and j, and a and b are counted by travel time on the road.8 trade elasticity parameters based on Deichmann Pj (1997). We then summarize market access at an MAi = ∑ t ij-q administrative level for each country by converting j � the market access results to an inverse distance where MAi is market access in location i, Pj is the weighted grid and taking the mean of the grid in population in location j, tij is travel time between the administrative level. The spatial distribution locations i and j, and q is a trade elasticity param- of the market access indicator is presented in eter. We then apply the most commonly used Map 2.5. (Appendix B provides further details of modified model as it is more relevant to coun- the construction of our market access indicator.) tries with geographically limited data on populated There are two key limitations to our model. locations.9 First, we only consider land transportation (e.g., road networks), thus potentially underestimating market access index in coastal areas, where sea access is available. Second, our model computes 8 Examples from the literature with similar market access include domestic market access with locations across Harris (1954); Hanson (2005); Emran and Shilpi (2012); Jedwab and Storeygard (2015); Berg, Blankespoor, and Selod (2016); and Don- borders not being taken into account. As a result, aldson and Hornbeck (2016). market access index of areas along a country’s 9 See Lall, Shalizi, and Deichmann (2004); Yoshida and Deichmann border may also be underestimated. (2009); and Ballon et al. (n.d.). Geography of Welfare—Three Building Blocks 13 Map 2.5  Concentration of High Market Access in the South, around Capitals and Economic Capitals a. Benin b. Burkina Faso c. Côte d’Ivoire d. Togo Source: Authors’ calculations based on data from DeLorme (2015). Two striking facts emerge. First, across all (Côte d’Ivoire), and Lomé (Togo). However, not all four countries, the North generally has lim- coastal areas are born equal, as shown by Côte ited access to market. Second, areas with high d’Ivoire, where the western coastal side of the market access cluster in the South, around the country does not enjoy the same levels of market capitals and the economic capitals—Cotonou access as the eastern side, at least not until it (Benin), Ouagadougou (Burkina Faso), Abidjan reaches closer to the border with Liberia. 14 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo Chapter 3 Spatial Disparities in Welfare and Poverty This chapter visually presents the geographical as irrigation systems for poor farmers, or infor- distribution of welfare and poverty and relates mation services such as mobile phone coverage. it to the three key elements of economic geog- The spatial distribution of poverty density, defined raphy: natural endowment, agglomeration, and as the number of poor people per square kilome- market access. As described in Chapter 2, spatial ter, and maps of current public services coverage disparities in welfare could be a natural outcome are critical for policy makers to decide whether a of the cycle of development: as a country devel- new service delivery program can be offered or ops, economic activity concentrates in certain an existing program can be expanded in a cost- regions to take advantage of economies of scale, effective way. If so, how many locations can the with these regions in turn attracting more people programs reach, and where are these locations to looking for job opportunities, which consequently be found? The coverage of such programs depends increases population density. With increasing heavily on the projected costs (e.g., upfront invest- returns to scale, these regions can continue to ment such as schools, piping for water connec- invest in access to markets and thus further rein- tions, electricity lines and poles, etc.), which in force their advantage. However, this development turn are largely determined by the density of users process makes it more difficult for poorer areas and the current status of public service coverage. to catch up. Finally, by assessing how the characteristics of the If economic development is inevitably uneven poor differ across space from food consumption within a country, where are the leading and lag- patterns to household demographics, we aim to ging areas located, and where do the majority of help governments quickly identify affected groups the poor live? These are among the questions we in cases of shock (e.g., rise in commodity prices, aim to answer in this chapter. Within the domain etc.) or policy reforms related to food products, of welfare and poverty, we focus on four important such as maize subsidies. Within a country, the dimensions: leading and lagging regions, poverty poor display distinct characteristics and face dif- measures (including poverty rates, poverty mass, ferent challenges in each location. For example, and poverty density), access to basic services, and factors with significant impact on the poor’s wel- profile of the poor. fare in the North, such as maize price, may play a lesser role than those in the South. Arguably, the stylized facts presented in this chap- ter are useful for policy makers for several reasons. To preview our main findings: First, they can guide budget allocation across 1. There is a large income gap between lead- administrative units by identifying which regions ing and lagging areas. This gap is highest in have fallen behind in terms of economic devel- Benin and lowest in Burkina Faso. opment, which regions have forged ahead, and more importantly, the magnitude of the income 2. Many of the leading areas have not reached gap between them. their full potential. In other words, many have not yet maximized the benefits of agglomera- Second and along the same lines, the geographi- tion economies (especially in Burkina Faso and cal targeting of programs designed to alleviate Côte d’Ivoire) or of market access. poverty can benefit from the identification of geo- graphical areas with high prevalence of poverty, or 3. Within a country, there is wide variation in poverty rates, defined as the share of population poverty rates, such that the North is mark- living below US$1.90 a day at 2011 PPP. In addition edly poorer than the South (except in Burkina to poverty rates, information on the number of the Faso). poor, or poverty mass, is handy when it comes to 4. The poverty mass—the number of poor cost estimates of a social policy targeted to the people—is highest in low-density areas (in poor, such as social safety net programs. Burkina Faso, Côte d’Ivoire, and Togo). This Third, public investments in service delivery pro- pattern suggests that the cost for service grams designed for the poor can be prioritized delivery programs to physically reach the poor accordingly. In this context, services may come could be relatively high, especially as access to in many forms and include social services such as public services such as improved toilets, piped primary education for all, economic services such water, and electricity differs greatly across 16 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo space, with the North having lower coverage than the South. Leading and Lagging Areas 5. In each country, there exist geographical As discussed in Chapter 2, the accumulation of pockets of poverty, which may be resis- wealth in one area but not in another may be a tant to policy-induced change. Related to natural outcome of the development path (made the main findings, #3 and #4 above, these plausible through the evolution of agglomeration areas are characterized by a combination of and expansion in market access). The question high poverty rates and low poverty densities. is therefore: What regions are benefiting from Therefore, the unit cost of a poverty-targeted the fruits of development and which are falling program could be extremely high in these behind? More importantly, how wide is the gap areas. Given budget constraints, the govern- between them? ment may not be able to reach this population In fact, there are only a few leading areas and group. many lagging ones (Map 3.1. We define a loca- 6. Looking beyond monetary poverty (i.e., tion as “leading” if per capita consumption for an US$1.90 a day at 2011 PPP), the quality of average household living there is higher than the life of the poor measured by the extent of national average. Out of 107 departments in Côte food intake diversification and housing con- d’Ivoire,10 only 19 are leading. This figure is 14 out ditions varies across space, with geographi- cal patterns mimicking those observed with 10 Although Côte d’Ivoire has 108 departments, the 2001 ENV survey monetary poverty. covers only 107 of them. Map 3.1  Cluster of Leading Areas in the South, around the Capitals and the Economic Capitals, or along Country Borders Sources: Authors’ calculations based on Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. Note: A location is defined as “leading” if consumption per capita for an average household living there is higher than the national average. Spatial Disparities in Welfare and Poverty 17 Figure 3.1  Large Wealth Gap between Leading and Lagging Areas Benin Burkina Faso Consumption per capital, $ PPP 2011 Consumption per capital, $ PPP 2011 2,000 1,500 1,500 1,000 1,000 500 500 0 0 go be li i vo ou ou m a et el o dj ob om og bi u ou o no ar m n ur ro C kp m to di ou nd H op So Lo ou rto Ka on e- uk Zo C m N Po C Bo Se Côte d’Ivoire Togo Consumption per capital, $ PPP 2011 Consumption per capital, $ PPP 2011 2,500 2,000 2,000 1,500 1,500 1,000 1,000 500 500 0 0 e la u yo o n ou a ou cs e fe um lo gl ja itt un re ol La ue lg eb pi ui id Bl G ng m fe O Si G Ab G Ak om ou Te D C e m Sources: Authors’ calculation based on Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo Lo QUIBB 2015. of 77 communes for Benin, 11 out of 45 provinces a large wealth gap between leading and lagging for Burkina Faso, and 9 out of 36 prefectures for locations. As shown in Figure 3.1 the difference in Togo.11 income between the top three leading areas and the bottom three lagging ones could be as high as The limited number of areas with per capita con- a factor of 7 (in the case of Benin). In this regard, sumption above the national average suggests Burkina Faso is the least unequal country, with a ratio of approximately 3.5 between the province 11 For Benin, leading communes are Abomey, Abomey-Calavi, with the highest income and the poorest one. Adjarra, Bohicon, Cotonou, Dassa-Zoumé, Houeyogbe, Natitingou, Ouèssé, Parakou, Porto-Novo, Sakété, Savé, and Sèmè-Kpodji. Lead- Surprisingly, many of these leading areas have ing provinces in Burkina Faso are Boulgou, Comoé, Houet, Kadiogo, not yet maximized the benefits of agglomeration Nahouri, Noumbiel, Oudalan, Poni, Sanmatenga, Séno, and Yagha. economies. This observation is especially clear in For Côte d’Ivoire, Abidjan is one of the leading departments, the rest being Abengourou, Aboisso, Adzopé, Bangolo, Bettié, Blolequin, Burkina Faso and Côte d’Ivoire (Figure 3.2). In a Bouaflé, Dabou, Duékoué, Gagnoa, Grand-Bassam, Guéyo, Guiglo, country with sparse population such as Burkina San-Pedro, Sikensi, Tabou, Yamoussoukro, and Zuénoula. For Togo, leading prefectures are Bassar, Cinkassé, Danyi, Golfe, Lacs, Lomé, Faso, a majority of the better-off provinces are Ogou, Tchaoudjo, and Vo. still located in ultra-remote areas. The fact that 18 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo Figure 3.2  Many Leading Areas in Low-density Locations Benin Burkina Faso log (consumption per capita) log (consumption per capita) 7.5 7.5 ultra-remote rural rural urban ultra-dense urban ultra-remote rural rural urban ultra-remote urban 7 7 6.5 6 6.5 5.5 6 2 4 6 8 10 2 3 4 5 6 log (population density) log (population density) leading region leading region Côte d’Ivoire Togo log (consumption per capita) log (consumption per capita) 8 8 ultra-remote rural rural urban ultra-dense urban ultra-remote rural urban ultra-dense urban 7.5 7.5 rural 7 7 6.5 6.5 6 6 2 4 6 8 3 4 5 6 7 8 9 log (population density) log (population density) leading region leading region Sources: Authors’ calculation based on Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. these provinces have not taken advantage of agglomeration economies, the nature of the eco- increasing returns to scale by urbanizing could nomic activities in this area, which consists mostly partially explain why the wealth gap between rich of informal trading with Nigeria, might attract and poor provinces is relatively low in Burkina large populations of poor migrants (Golub, 2012). Faso. We notice a similar pattern in Côte d’Ivoire, Similarly, we find a mixed pattern between where nearly half of the leading departments are leading and lagging areas and market access. located in rural areas. However, some of the lead- Although a region with higher per capita consump- ing departments in the country host large-scale tion is often shown to have better market access, farmers, who need land of considerable size for this pattern does not always hold. Figure 3.3 illus- their operations, thus explaining their low-density trates how many leading administrative units in locations. fact have low market access (defined as having It is important to point out a mixed picture in market access below the average value across all Benin. While some leading communes have low four countries) and vice versa. As mentioned in population density, many lagging ones are situ- Chapter 2, a limitation of our market access index ated in either urban or ultra-dense urban areas. results from underestimates of market access Why, therefore, do these locations, which could values in administrative units along a country’s enjoy the benefits of agglomeration externalities, border and along the coast. This could explain why remain poor? In fact, these dense but lagging com- some leading administrative units do not have munes cluster in the South near the coast of Benin high market access. (Map 2.3 and Map 3.1). While the poor might enjoy Spatial Disparities in Welfare and Poverty 19 Figure 3.3  Low Market Access in Many Leading Areas Benin Burkina Faso log (consumption per capita) 7.5 log (consumption per capita) Low MA 7.5 Low MA High MA High MA 7.0 7.0 6.5 6 6.5 5.5 6 6 8 10 12 14 0 5 10 15 log (market access) log (market access) leading region leading region Côte d’Ivoire Togo High MA log (consumption per capita) 8 7.5 Low MA log (consumption per capita) Low MA High MA 7.5 7.0 7 6.5 6.5 6 6 0 5 10 15 6 8 10 12 14 log (market access) log (market access) leading region leading region Sources: Authors’ calculation based on Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. Poverty Rates, Poverty the North, which is the case in neighboring coun- tries such as Mali and Niger. Mass, and Poverty Density Not surprisingly, leading areas have the lowest The previous section discussed leading and lag- poverty rates and are clustered in the South, ging areas in terms of income.12 We now turn to an around the capitals and the economic capi- equally pressing concern: Which areas experience tals, and along the country’s border (Map 3.3). widespread poverty incidence and which ones do This pattern of poverty is consistent with the not? Here we use the US$1.90-a-day poverty line geographical distribution of leading and lagging at 2011 PPP to define poverty. regions described previously. Here again, we observe a tale of two regions: However, what is alarming is the large varia- North vs. South (Map 3.2). Within each of the tion in poverty incidence between leading three coastal countries—Benin, Côte d’Ivoire, and and lagging areas. In Benin, poverty rates can Togo—the North, corresponding to the two least vary between 20 percent in Cotonou and nearly favorable agro-ecological zones (Zones 1 and 2) is 100 percent in the three most lagging communes markedly poorer than the South, which comprises (Cobli, Copargo, and Boukoumbé). A similar range the two most favorable zones (Zones 3 and 4). In is observed in Togo, or between 15 percent in the Togo particularly, poverty in the far North may be top three prefectures (Golfe, Lacs, and Lomé) and more than three times as high as in the far South. above 90 percent in the bottom three (Tandjoaré, However, the pattern of poverty is reversed in Akebou, and Doufelgou). For Burkina Faso, the Burkina Faso. One possible explanation is the pos- figure ranges from about 10 percent in the rich- session of livestock among Burkina inhabitants in est provinces (Noumbiel, Nahouri, and Kadiogo) to above 80 percent in the most disadvantaged ones (Komandjoan, Zondama, and Sourou). Côte 12 Measured as per capita consumption. d’Ivoire tells the same story, with poverty rates 20 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo Map 3.2  In Three out of Four Countries, the North Is Remarkably Poorer than the South Sources: Authors’ calculations based on Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. Map 3.3  Lower Poverty Rates in Areas around the Capitals and the Economic Capitals, and along the Country Border Sources: Authors’ calculations based on Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. Spatial Disparities in Welfare and Poverty 21 between 8 percent in the top three departments At subnational levels, some leading areas in (Guéyo, Abidjan, and Tabou) and over 80 percent Benin and Côte d’Ivoire host the highest number in the bottom three (Tengrela, Sipilou, and Oumé). of poor people.13 These are Abomey-Calavi com- Figure C.1 in Appendix C provides details of poverty mune, suburban Cotonou in Benin, and Abidjan rates at subnational levels for each country. department in Côte d’Ivoire. In contrast, in Burkina Faso and Togo, the largest poor population is con- A relevant consideration for policy makers is not centrated in two of the most disadvantaged areas only the issue of where poverty rates are high but in terms of income: Yatenga province in Burkina also where the poverty mass is located. Here, we Faso, and Oti prefecture in Togo. define poverty mass as the number of poor people. A location with a lower poverty rate does not nec- Nevertheless, most of the poor still reside in essarily imply that it has fewer poor people when either ultra-remote rural or rural areas, where population is taken into account. density is lower than 150 people per kilometer (Figure 3.4).14 This strong pattern seen in Burkina In fact, relatively better-off locations, such as Faso, Côte d’Ivoire, and Togo, where at least three the capitals or the economic capitals of coun- out of four poor people live in low-density loca- tries and areas in the South, have high poverty tions, suggests that the cost of physically reach- mass. In Benin and Côte d’Ivoire, nearly half of the ing the poor could be relatively high in these three poor congregate in Zone 4 (the South). Similarly, countries. in Burkina Faso, about half of the poor population lives in Zone 2 in and around the capital. Mean- while, Zone 3 in Togo is home to about 40 percent 13 Compared to other administrative units in the same country. of the country’s poor. 14 Following our agglomeration classification in Chapter 2. Figure 3.4  Majority of the Poor Living in Low-density Areas Poverty rates ($1.90/day, PPP 2011) BEN BFA ultra-remote rural ultra-remote rural rural rural urban ultra-dense urban ultra-dense urban CIV TGO ultra-remote rural ultra-remote rural rural rural urban urban ultra-dense urban ultra-dense urban 0 .2 .4 .6 .8 0 .2 .4 .6 .8 Percent of population Number of poor people by agglomeration BEN BFA CIV TGO ultra-remote rural rural urban ultra-dense urban Sources: Authors’ calculations based on Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. 22 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo Map 3.4  High Poverty Density in and around the Capitals and the Economic Capitals, and in the South Sources: Authors’ calculations based on Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. The final—though no less important—dimension poverty density, is strikingly high, or nearly 200 of poverty discussed in this section is poverty den- times the number of poor people. This ratio is 120 sity. Similar to population density, poverty density for Togo, 73 for Côte d’Ivoire, and 31 for Burkina is defined as the number of poor people per square Faso. Figure C.2 in Appendix C lists poverty density kilometer (Map 3.4). Not surprisingly, the highest for each administrative unit in the four countries density of the poor is found in the capitals or of interest. the economic capitals. Across countries, Coto- nou commune in Benin is the densest, at nearly 1,700 poor people per square kilometer, Lomé commune in Togo follows, with poverty density Access to Services of 1,500, and Abidjan department in Côte d’Ivoire Following our discussion of how density of users houses about 200 poor people per square kilo- could be one of the decisive factors affecting meter. In sparsely populated Burkina Faso, where service delivery coverage, we now examine dis- poverty density is relatively low, even the densest parities in the provision of services across space. province—Zondoma—has only 78 poor people per In this section, we focus on four service delivery square kilometer. programs available in our data sets, namely access It is important to note stark variation in pov- to electricity, piped water, improved toilet facili- erty density across administrative units, which ties, and mobile phone coverage.15 affects the cost of various government pro- We note that the share of the population having grams delivered to each location. In Benin, the access to electricity, piped water, and improved difference between Cotonou commune, with the highest number of poor people per square kilo- meter, and Karimama commune, with the lowest 15 Improved toilet facilities include flush toilets, ventilated improved pit latrines, composting latrines, and pit latrines. Spatial Disparities in Welfare and Poverty 23 Figure 3.5  Large Gaps in Public Service Coverage between the Most Sparsely and Most Densely Populated Areas BEN BFA 1 .8 .6 .4 .2 0 electricity improved mobile piped electricity improved mobile piped toilets phone water toilets phone water CIV TGO 1 .8 .6 .4 .2 0 electricity improved mobile piped electricity improved mobile piped toilets phone water toilets phone water ultra-remote rural rural urban ultra-dense urban Sources: Authors’ calculations based on Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. toilets increases with population density, with a national income nearly twice that of each of the the highest coverage observed in ultra-dense other three countries, takes the lead, with Burkina urban locations while the lowest is seen in ultra- Faso trailing far behind. In Côte d’Ivoire, one-third remote rural areas (Figure 3.5). This pattern is of the population has access to piped water, two- consistent with recent research that shows a close thirds use electricity in their house, and three- correlation between access to basic services and fourths have access to an improved toilet facility. population density (Gollin, Kirchberger, and Laga- For Burkina Faso, these figures are 11 percent, kos, 2016). Across agglomerations and countries, 47 percent, and 15 percent, respectively. mobile phone coverage is homogenously high by Within country, we observe a similar tale of developing country standards: on average, at least two regions, as discussed in the previous sec- 7 out of every 10 people in the subregion have a tions. While nationwide mobile phone coverage is mobile phone. This uniformly wide coverage could more or less equally distributed, coverage of piped be explained by the fact that mobile phone net- water, improved toilets, and electricity appears works require relatively low costs to reach end- to be divided into two regions (Map 3.5). In Benin, users compared to other service delivery programs Côte d’Ivoire, and Togo, the dividing line for the such as electricity or water. coverage of piped water, improved toilets, and However, there are wide differences in access electricity lies between the North (Zones 1 and 2) to services (piped water, improved toilets, and and the South (Zones 3 and 4), with a larger share electricity) across countries. Côte d’Ivoire, with of the population in the South having access to 24 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo Map 3.5  High Variation in Access to Public Services across Space (except cellphone coverage) Sources: Authors’ calculations based on Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. public services. In Burkina Faso, the division is Cotonou, is at the other end of the spectrum, with between the Center South (Zones 2 and 4) and more than 80 percent of residents having such the rest of the country. access. A similar story is observed in Côte d’Ivoire, where one of the poorest and most sparsely popu- At the subnational level, we document a wide lated areas, Sipilou department, shows less than geographical disparity in access to services, 15 percent of its population having electricity, which goes hand in hand with spatial differ- piped water, and improved toilets compared to ences in population density and poverty rates almost 100 percent of the population in the eco- (Map 3.3 and Map 3.6). Understandably, areas nomic capital, Abidjan. Togo shares the same pat- characterized by a combination of high poverty tern, with nearly no one in Blitta prefecture having incidence and low population density tend to have electricity or an improved toilet at home, while low services coverage (except for mobile phone over 90 percent of the residents of the capital, networks). The variation in services coverage Lomé, do. In Burkina Faso, Loroum province pro- can be large across space. In Benin, for example, vides electricity and piped water to almost none merely 2 percent of the population in Karimama of its inhabitants, while this figure is above 50 per- commune, one of the most sparsely populated cent for people living in Kadiogo province, where and poorest areas, have access to electricity and the capital, Ougadougou, is located. improved toilets. Meanwhile, the economic capital, Spatial Disparities in Welfare and Poverty 25 Map 3.6  Low Service Coverage in Areas with High Poverty Incidence Sources: Authors’ calculations based on Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. Quality of Life and the composition of the food basket, and housing conditions, as the poor spend a large share of their Characteristics income on food, and their house (and land) may be of Poor People their most valuable asset. In this section, we focus on housing because it is particularly relevant for Given that the previous sections show evidence poor people living in urban areas. Finally, we pro- of vast spatial disparities in poverty and access vide a snapshot of poor households’ demographics. to services, we would expect the quality of life for Our first measure of quality of life is the diver- poor people to vary greatly as well. While people sification of food intake. Keeping other factors may be considered poor based on monetary stan- constant, it is arguable that individuals are better dards (i.e., living below US$1.90 a day at 2011 PPP), off when they have a wider range of options for their living standards and demographic character- food items to be consumed, which enriches their istics may differ across space and reflect condi- diet and taste. We use the Herfindahl Index (HI), tions in the location where they live. To examine also known as the Hirschman Index or Hirschman- the poor’s quality of life, we explore the spatial Herfindahl Index as an inverse measure of variety differences in two important elements of daily life: 26 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo Map 3.7  Higher Diversity in Food Consumption Basket and Lower Food Share from Own Production in the South (a) (b)     Sources: Authors’ calculations based on Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. in food consumption.16 The HI ranges from 1/n to or live in agglomerations (Krugman, 1996; Fujita, 1, and reaches a maximum value of 1 if the share of Krugman, and Venables, 1999). consumption is entirely concentrated on a single Next, we examine the types of food that make up food item. In other words, the HI measures diver- the highest share of the poor’s food consumption. sity, where the higher the value of the index, the We first calculate the share of the food budget for lower the diversity (Lee and Brown, 1989). each item for each household in order to identify Map 3.7 suggests that in all countries, people the most commonly consumed food products. We in the South enjoy a more diverse food basket then estimate the share of the poor population than their fellow citizens living in the rest of the consuming these top-ranked food items. Map 3.8 country, with a lighter color indicating a lower HI.17 presents our results. Map 3.7 provides a complementary story: people In terms of products, the poor consume simi- in the South tend to consume less from their own lar food items throughout Benin, even though production (from a limited number of crop prod- the four agro-ecological zones reveal different ucts). As Chapter 2 pointed out, southern areas shares of the poor population consuming each generally have better market access and higher product. For example, a larger share of the poor population densities. Thus, their inhabitants rely have corn in their diet in Zone 4 than in the rest less on subsistence food production and more on of the country. However, corn is the top-ranked tradable food products. These observations are food item for poor people overall. In Zones 1, 2, consistent with the literature on taste for variety: and 3, the next-ranked food commodity is yam, people have more choices (of food intake, jobs, and while it is local rice for those living in the coastal so on) when they have better access to markets area (Zone 4). Burkina Faso exhibits a distinct pattern, with food consumption for most poor people in the North (Zones 1 and 2) consisting of 16 The HI is calculated as the sum of squared food shares: HIh = sorghum and millet, while the food items for those Sn s 2, where the HI of household h is the sum of the budget shares i=1 ih in the South (Zones 3 and 4) include corn and s of each individual food item i consumed in household h. The HI ranges from 1/n to 1. rice. In Côte d’Ivoire, most of the poor in Zone 1 17 We also note that food diversity measured by the HI should not consume yam, while the poor in the rest of the be compared across Benin, Burkina Faso, Côte d’Ivoire, and Togo country eat local rice. Those living in the coastal because the number and types of food commodities examined in household consumption surveys are not the same across countries. South (Zone 4) also have imported rice in their Spatial Disparities in Welfare and Poverty 27 Map 3.8  Variation in Key Food Consumed across Space Benin Burkina Faso Togo Côte d’Ivoire Sources: Authors’ calculations based on Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. food diet. Interestingly, in Togo, poor inhabitants of the poor population having concrete or tile roofs of the northernmost areas (Zone 1) consume or cement or brick walls increases significantly corn and local rice. Meanwhile, those living in the from the North (Zone 1) to the South (Zone 4). other three zones add dried fish and imported In all countries, almost no one in Zone 1 lives in a rice to their diet. Moreover, the share of the poor house with either concrete walls or tile roof. How- population stating dried fish as their most com- ever, this share of the poor population in Zone 4 monly consumed food product increases as we can be as high as 14 percent, as in the case of get closer to the coastal areas. Togo. Similarly, the difference between Zone 1 and Zone 4 in terms of share of the poor population Not only do the poor in the South enjoy a more with cement or brick walls can be from 2 times in diverse food intake, they also have better hous- Benin to nearly 10 times in Togo. We observe the ing conditions. Table 3.1 illustrates how the share 28 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo Table 3.1  Better Housing Conditions for Poor Households in Urban Areas or in Favorable Agroecological Zones Agroecological Zones Agglomeration Types Ultra- Ultra- remote dense Zone 1 Zone 2 Zone 3 Zone 4 Rural Rural Urban Urban House ownership  Benin 0.939 0.952 0.874 0.880 0.924 0.929 0.942 0.842   Burkina Faso n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.   Côte d’Ivoire 0.608 0.784 0.690 0.586 0.777 0.667 0.633 0.358  Togo 0.496 0.626 0.571 0.238 0.715 0.517 0.371 0.223 Concrete roof                  Benin 0.006 0.008 0.010 0.028 0.006 0.007 0.025 0.032   Burkina Faso 0.000 0.003 0.003 0.002 0.001 0.003 n.a 0.015   Côte d’Ivoire 0.011 0.002 0.007 0.007 0.008 0.004 0.011 0.014  Togo 0.003 0.004 0.001 0.138 0.000 0.004 0.050 0.187 Concrete walls                  Benin 0.19 0.40 0.48 0.45 0.33 0.42 0.26 0.55   Burkina Faso 0.02 0.05 0.06 0.12 0.05 0.08 n.a. 0.16   Côte d’Ivoire 0.59 0.54 0.39 0.50 0.41 0.46 0.56 0.78  Togo 0.06 0.19 0.21 0.62 0.09 0.16 0.28 0.80 Sources: Authors’ calculations using Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. same pattern across agglomeration types, where to 10 members, nearly twice as many as the poor people living in ultra-dense urban areas have number of their coastal neighbors, Benin, Côte significantly more chances to live in a house with d’Ivoire, and Togo. Within country, poor house- either concrete or tile roof or cement or brick walls. holds in the most disadvantaged agroecological zone (Zone 1) or in ultra-remote rural areas tend However, poor people in the South and in ultra- to have larger household size and a higher share of dense urban areas are less likely to own their young dependents than their counterparts living house. One plausible reason could be the relatively in more favorable zones (Zones 2, 3, and 4) or in high real estate prices in urban areas, such that denser population areas. for the same levels of income, a poor person in the city cannot afford to buy or build a house com- When we look at the demographic character- pared to a similarly poor person living in a rural istics of heads of poor households, a pattern area. Another possible explanation could be that emerges: very few poor households are female- the poor population in urban areas may comprise headed in ultra-remote rural areas or in Zone 1 migrant workers who need temporary housing. (Table 3.2). The share of poor households headed by a female increases for those living in better-off Lastly, we consider how poor households’ demo- locations, such as Zone 4 or ultra-dense areas. graphics vary across space (Table 3.2). In these This observation echoes results of recent studies predominantly agricultural economies, the number of female-headed households, which found that of able-bodied household members plays a criti- in poor areas characterized by a scarcity of jobs cal role in determining the scope of households’ or with economic activity heavily dependent on economic activity and, consequently, their pov- agriculture, widowed or divorced women cannot erty status. As expected, in areas where natural afford to be on their own and therefore tend to circumstances are harsh and labor productivity merge with other households (e.g., sons’, broth- is low, households need more members to gener- ers’, or through remarriage, etc.). Meanwhile, in ate income. In fact, a typically poor household in relatively richer areas, where more employment landlocked Burkina Faso, on average, has close Spatial Disparities in Welfare and Poverty 29 Table 3.2  Fewer Family Members, Lower Dependency Rates, More Likely to Be Female-Headed Households, and Less Likely to Have No Education among Poor Households in Urban Areas and Favorable Zones Agroecological Zones Agglomeration Types Ultra- Ultra- remote dense Zone 1 Zone 2 Zone 3 Zone 4 Rural Rural Urban Urban Poor households Household size  Benin 6.0 6.1 5.4 4.7 6.05 5.50 4.69 4.70   Burkina Faso 10.0 10.7 9.1 8.7 9.60 10.13 n.a. 8.19   Côte d’Ivoire 5.6 6.2 4.4 5.2 5.47 5.00 4.79 5.96  Togo 6.9 5.9 5.6 5.1 6.46 6.05 5.51 5.33 Number of working-age members   Benin 2.45 2.48 2.41 2.14 2.52 2.35 2.15 2.18   Burkina Faso 4.05 4.49 3.54 3.81 4.01 4.24 3.91   Côte d’Ivoire 2.42 2.78 2.15 2.51 2.43 2.33 2.26 3.09  Togo 3.00 2.64 2.55 2.43 2.86 2.66 2.50 2.72 Share of household members under 15          Benin 0.53 0.52 0.48 0.47 0.51 0.50 0.47 0.47   Burkina Faso 0.53 0.51 0.56 0.51 0.53 0.51 n.a. 0.46   Côte d’Ivoire 0.50 0.49 0.43 0.45 0.48 0.45 0.45 0.46  Togo 0.51 0.46 0.46 0.40 0.48 0.48 0.42 0.40 Head of poor households Male  Benin 0.92 0.90 0.80 0.76 0.89 0.83 0.78 0.75   Burkina Faso 0.92 0.89 0.90 0.86 0.90 0.87 n.a. 0.86   Côte d’Ivoire 0.69 0.87 0.83 0.82 0.81 0.83 0.79 0.76  Togo 0.85 0.69 0.68 0.64 0.74 0.72 0.68 0.76 Married  Benin 0.89 0.86 0.82 0.82 0.87 0.84 0.82 0.82   Burkina Faso 0.94 0.88 0.93 0.87 0.88 0.90 n.a. 0.74   Côte d’Ivoire 0.62 0.89 0.64 0.71 0.77 0.69 0.71 0.64  Togo 0.81 0.71 0.82 0.69 0.83 0.77 0.75 0.74 No education          Benin 0.93 0.86 0.76 0.63 0.88 0.81 0.69 0.61   Burkina Faso 0.95 0.92 0.93 0.88 0.90 0.93 n.a. 0.87   Côte d’Ivoire 0.75 0.88 0.62 0.68 0.81 0.68 0.71 0.56  Togo 0.66 0.45 0.40 0.33 0.36 0.54 0.45 0.19 Sources: Authors’ calculations using Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. 30 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo opportunities are available, we observe more This suggests two conclusions: poor people with female-headed households (van de Walle, 2013; some education may migrate to better-off areas, 2015). This pattern is consistent with the fact that or there may not be enough jobs to absorb the the share of married heads of poor households is poor labor force with some education in cities. This lower in resource endowed or dense areas. It is problem is more pronounced in Togo, where only interesting to note a smaller share of poor house- 19 percent of poor households in ultra-dense cities hold heads with no education in more favorable have no education. This means that although most locations such as Zone 4 or ultra-dense cities. of them have some education, they are still poor. Spatial Disparities in Welfare and Poverty 31 Chapter 4 Geographical Differences in Agricultural Activity In this subregion as around the world, agriculture sector as a share of GDP and employment, it fol- is a cornerstone of the economy. The agriculture lows that the agriculture sector can—and should— sector forms a large part of the economy and play a potentially important role in growth and retains great potential to foster inclusive growth. poverty reduction. In fact, this crucial link between improvements to Beyond its implications for food security, a focus the agriculture sector and reductions in poverty on agriculture and agricultural productivity is cru- is often observed. cial to poverty reduction for other reasons. First, In 2015, the agriculture sector accounted for it makes sense to improve the productivity of a roughly a third of GDP for the countries in this sector on which most of the poor rely. Among the subregion, or 25, 34, 20, and 41 percent for Benin, set of countries in this subregion, compared to Burkina Faso, Côte d’Ivoire, and Togo, respectively. the nonpoor, more of the poor are engaged in agri- The size of the agriculture sector in these coun- culture. In Togo, for example, among the nonpoor, tries is comparable to that in low-income coun- around 40 percent engage in agriculture, while tries, where 31 percent of GDP is attributed to among the poor, close to 80 percent do so. Prior agriculture, and it is even larger when compared to work suggests that agricultural growth reduces Sub-Saharan Africa (SSA), where only 18 percent poverty by an amount three times greater than of GDP is attributed to agriculture, and especially growth in other sectors (Christiaensen, Demery, when compared to the rest of the world, where and Kuhl, 2011). Second, improving agricultural only 5 percent of GDP is attributed to agriculture. productivity can foster structural transformation and manage the urban transition by increasing The agriculture sector also provides incomes and incomes and promoting nonfarm jobs and enable employment to the vast majority of the population. people to move out of agriculture over time (Gollin, For example, in Côte d’Ivoire, agriculture employs Lagakos, and Waugh, 2014; McMillan and Hartt- 67 percent of households and subsistence farming gen, 2014). employs 85 percent of the population. In Burkina Faso, agriculture is the main economic activity This chapter explores the geography of poverty for 70 percent of households, employing around in this subregion by examining the geography of 80 percent of the working population. Similarly, agriculture. It focuses on the relationship between in Benin, agriculture accounts for 70 percent of agroecological zones and key aspects of agricul- employment. Given the size of the agriculture ture, including productivity in crop production, use Figure 4.1  Percentage of Population Engaged in Agriculture, by Poor and Nonpoor 100 94.4 80 78.7 78.6 77.9 % of individuals in agriculture 67.5 66.5 60.3 62.1 60 58.2 55.4 51.0 51.4 44.7 43.0 41.1 40 20 0 CIV BFA TGO BEN Subregion All Nonpoor Poor Sources: Authors’ calculations using Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. Geographical Differences in Agricultural Activity 33 of agricultural inputs and land tenure security, and Compared to the nonpoor, a larger propor- sale of output. The agricultural and land modules tion of the poor are engaged in agriculture. contained in the most recent household surveys In Benin, for example, while 40 percent of the are used for each of the four countries, thus giv- nonpoor engage in agriculture, 80 percent of ing an opportunity to make further connections the poor do so. In Burkina Faso, about 95 per- between agricultural activity and poverty at the cent of the poor engage in agriculture. How- household level. ever, among those in agriculture, those with higher maize yields show higher per capita A key distinguishing factor in our analysis of agri- expenditure. Thus, although there is a high cultural activity is precisely that we rely on the proportion of poor in agriculture, improving agriculture and land modules contained in house- farm yields can improve their conditions. hold surveys as opposed to administrative data on agricultural production. Thus, although this 3. The agriculture sector varies across coun- is unlikely to adequately capture the agriculture tries, with a much more developed sector sector of a country, it is more likely to capture in Côte  d’Ivoire and a much weaker sector the sector from the perspective of households in Burkina Faso. In Côte d’Ivoire, the crops and is therefore more likely to be biased toward that are most prevalent (i.e., planted by most smallholder agriculture as opposed to large-scale individuals) are a combination of cash crops commercial farming. However, we should warn (crops produced for commercial value) such that despite the sector’s importance in this sub- as cacao and cashew, and food crops such region, as, for example, in northern Burkina Faso, as yam and maize, while food crops are domi- livestock and transhumance farming is beyond the nant in Burkina Faso, Togo, and Benin. Com- scope of this study. Our findings should therefore paring Burkina Faso to Togo, there is more be interpreted with this limitation in mind. market exchange of food crops in Togo. For example, while 10 percent of sorghum grow- More broadly, mapping productivity, input use, ers in Burkina Faso sell some of their crop, and output sales across agroecological zones will 57 percent of sorghum growers in Togo do so. enable policy makers to better target agricultural Moreover, the zone with the lowest maize yield interventions aimed at increasing growth and in Côte d’Ivoire has higher yields than the zone reducing poverty. with the highest maize yield in each of the We summarize the key findings in this chapter other countries. Although cotton production as follows: exists across all countries, its intensity var- 1. Relative to the service and industrial sec- ies within countries. In agroecological zones, tors, the agriculture sector employs a larger where cotton is not produced to any great share of individuals in all zones, except in extent, other cash crops come to the fore, coastal Zone 4 in Côte d’Ivoire and Togo. The except in Burkina Faso. dominance of agriculture is most pronounced 4. Within countries, agricultural performance in Burkina Faso, where the share of employ- varies across zones in the sense of higher ment in agriculture is 95 and 92 percent in yields or higher revenues. Better agroeco- Zones 1 and 3, respectively. Across zones, logically endowed zones do not necessarily employment in industry is lowest (under outperform other zones. Zones with a key 10 percent in many zones), even though vari- cash crop, particularly cotton, tend to have ation in employment in industry is lower in higher revenues. In Côte d’Ivoire, Zone 1 is Côte d’Ivoire, suggesting that there is a more largely disadvantaged, with low yields and equal geographic spread of opportunities for low revenues, while Zone 2, which produces employment in industry there than elsewhere cotton, has the highest revenues from crop in the subregion. sales. Similarly, in Burkina Faso, Zone 4, which 2. There is a link between poverty and involve- produces cotton but is also the better agro- ment in agriculture as well as between ecologically endowed zone, has the highest expenditure and agricultural productivity. revenues from crop sales, while Zone 1 has the This suggests that the geography of poverty is lowest revenues from sales but better maize intertwined with the geography of agriculture. yields than Zones 2 and 3. In Togo, Zone 2, 34 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo which spends a great deal on inputs and also in Zone 4 (39 percent), whereas it is highest in has higher maize yields, has lower revenues Zone 2 (77 percent). In Togo, there is a notice- from crop sales, while Zone 1 has the highest ably lower share of employment in agriculture in revenues from sales. In Benin, cotton yields Zone 2, where only 38 percent are employed in are highest in Zone 1, while maize yields are agriculture. Additionally, in Togo, only 10 percent highest in Zone 2. are employed in agriculture in Zone 4, the low- est share of agriculture in employment across all In Appendix D, a table is provided which provides zones in all four countries. Similarly, Côte d’Ivoire a brief summary of each zone in each country. shows large shares of employment in agriculture in Zone 1 (54 percent), Zone 2 (56 percent), and Zone 3 (59 percent), whereas in Zone 4, agricul- Employment in Agriculture ture employs only a little under 40 percent of individuals. Across all countries, the agriculture sector The industrial sector employs the smallest employs a larger share of the population in all share across agroecological zones, especially agro-ecological zones, except in Zone 4 of Côte in Burkina  Faso. Variation across zones in the d’Ivoire and Togo, where the services sector industrial sector’s share of employment is employs a larger share. The coastal Zone 4 of smallest in Côte d’Ivoire, suggesting a better Benin may be less similar to Zone 4 in Côte d’Ivoire geographic spread of industry across the coun- and Togo at least partly because Zone 4 in Benin try. In Burkina Faso, there are visible—though still presents much greater diversity as it consolidates only small—shares of industrial sector employ- information on three different livelihood zones. ment in Zones 2 and 4, with a share under 10 per- Note, however, that while agriculture still employs cent. Within Togo, Zone 1 has the smallest share a larger share in Zone 4 in Benin, the industrial and employed in the industrial sector (8 percent), while services sectors combined employ more than the Zone 4 has the largest share (47 percent). Within agriculture sector. Benin, the pattern is similar, with Zones 1 and 2 Across countries, Zones 1 and 3 in Burkina Faso having the smallest share (under 10 percent), and have distinguishably higher shares of employ- Zone 4 having the largest share (23 percent). In ment in agriculture, while Zone 4 in Togo has Côte d’Ivoire, variation across sectors in the share a distinguishably lower share. In Burkina Faso, of the population employed in the industrial sector the share of employment in agriculture is 95 and is smaller, while the share of industry is equally 92 percent in Zones 1 and 3, respectively. In Benin, substantial in Zones 2 and 4, both being close to the share of employment in agriculture is lowest a 20 percent share of employment in industry. Map 4.1  Employment in Agriculture Relative to Other Sectors Sources: Authors’ calculations based on Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. Geographical Differences in Agricultural Activity 35 Within zones, those employed in agriculture individuals engaged in agriculture. We use this are evenly spread. In locations where there are as an indicator of which crops are important in agglomerations, those people are more likely to the sense that more individuals engaged in agri- be employed in industry and services. culture rely on such crops. Thus, the focus is on crops grown by most people as opposed to crops that employ the most wage workers, take up more land, show higher output, or have the most value. Agricultural Productivity The choice to focus on number of people growing a crop is largely due to data constraints. Neverthe- While the agriculture sector faces various chal- less, this choice of focus also provides an opportu- lenges and goals, enhancing agricultural produc- nity to understand the agriculture sector from the tivity is paramount for both poverty reduction and smallholder angle as opposed to an understanding economic growth and thus central to agriculture driven by large farming activity. We observe that policy objectives. Growth in the agriculture sector both across and within countries, there is a great is at least three times more effective in reducing deal of variation in the types of crops grown. poverty than growth originating in the rest of the economy (Ligon and Sadoulet, 2007). Particularly In Côte d’Ivoire, the crops produced by most in Sub-Saharan Africa, leveraging agriculture- people are a combination of food crops and cash led growth requires a productivity revolution for crops. Half of the population grows cacao, fol- small­ holder farmers (World Bank, 2008). Glob- lowed by yam (29 percent), maize (21 percent), ally, a 1 percent increase in yields is associated and cashew (18 percent). While cacao and cashew with a 0.91 percent decrease in the percentage are largely sold in the market, only 20 percent of of the poor population (Irz et al., 2001). However, those that plant yam and 35 percent of those that it is worth noting that relative to other regions, plant maize sell some of those crops. both yields and poverty have changed least in Sub- In Burkina Faso, Togo, and Benin, the crops pro- Saharan Africa (De Janvry and Sadoulet, 2010; duced by most people are food crops. However, World Bank, 2008). there seems to be more market exchange of For each of the countries in this subregion, cereal crops in Togo than in Burkina Faso. In Burkina yields were lower than the world average, even Faso, the crops planted by most are sorghum though Côte d’Ivoire clearly outperformed the (77 percent), mil (55 percent), and maize (54 per- other three countries. In 2014, the world average cent). Cash crops, such as cotton and sesame, are for cereal yields was 3,886 kg/hectare, while it grown by only 15 percent and 29 percent, respec- was only 1,460 kg/hectare in Benin, 1,226 kg/­ tively, of those working in agriculture. In Togo, the hectare in Burkina Faso, 1,146 kg/hectare in Togo, most commonly produced food crops are maize and 3,254 kg/hectare in Côte d’Ivoire. From 2010 (93 percent), cowpeas (57 percent), and sorghum to 2014, Côte d’Ivoire saw a high increase in cereal (41 percent). Interestingly, there seems to be more yields (93 percent), whereas Togo had a minimal market exchange of food crops in Togo than in increase (8 percent). For the same period, cereal Burkina Faso. For example, of those who plant yields for Benin and Burkina Faso increased by 32 maize in Burkina Faso, 20 percent sell some of their and 43 percent, respectively (World Bank, 2017). crop, whereas in Togo, 45 percent do so. Of those that plant sorghum in Burkina Faso, only 10 per- In this section, the major crops of each country cent sell some of their sorghum crop, whereas in are first summarized, comparing crops grown by Togo, 57 percent do so. Much like Burkina Faso and the poor versus the nonpoor. Then, in an effort to Togo, Benin has most of its agricultural popula- understand variation in productivity within each tion producing some cereals (93 percent). With country, yields for a representative food crop respect to cash crops, about 26 percent of those (maize) and cash crops are mapped out across engaged in agriculture in Benin produce palm oil, agroecological zones. and 23 percent produce cotton. Major Crops Overall, the crops produced by the poor are very similar to those produced by the nonpoor. Simi- As a glimpse into the agriculture sector for each lar proportions of the poor and nonpoor grow key country, we look at which crops are grown by cash crops. However, there are distinctions in the 36 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo Box 4.1  Summary of the agriculture sector based on various World Bank documents (Project Information Document [PID], Project Appraisal Document [PAD], and Systematic Country Diagnostic [SCD]) This inset provides a brief background on the agri- Community Based Rural Development, PID, 2012; culture sector in each country to demonstrate that Additional Financing for Agricultural Diversification agricultural productivity, among other aspects, is and Market Development PID, 2014). central to the agenda. In Togo, one of the pathways toward inclusive growth In Côte d’Ivoire, despite higher productivity relative is the transformation of agriculture toward more pro- to the other three countries, the lack of agricultural ductive, higher-value, and sustainable smallholder development and diversification and of structural and commercial production. Despite a comparative transformation into agro-business are policy issues advantage in agriculture, Togo has been unable to to which the country’s overall poor performance can raise productivity consistently in the sector, including be attributed. Inclusive growth can be achieved at through diversification into higher value-added prod- least partially by developing the agriculture sector ucts. Agriculture is mainly of the subsistence type, through enhanced productivity and developing agro- and has seen a deterioration in the performance of business and non–agro-business sectors. Despite its the main export crops: cotton, coffee, and cacao. Togo importance to the economy, the agriculture sector has begun to address some of the constraints that has had only a modest impact on income growth affect the sector by recently adopting a plan aiming and poverty reduction in rural areas as the sector to reform the fertilizer subsidy program (Togo SCD, is characterized by low and unstable value addition. 2016). For example, while Côte d’Ivoire is the world’s largest In Benin, the government strategy for the agricul- exporter of cacao, this crop has seen declining pro- ture sector has been to improve productivity and ductivity, and cacao farmers are falling into poverty. strengthen diversification. Several challenges plague Other key cash crops include rubber, palm oil, cotton, the sector. One of these is the reliance on a limited and cashew nuts (Côte d’Ivoire SCD, 2015; Agriculture number of agricultural products, with most public Sector Support Project PAD, 2013). resources concentrated on a single crop, cotton, In Burkina Faso, the lack of opportunities for rural which accounts for 25–40 percent of GDP. Another populations is due to limited productivity gains in challenge is the low level of both productivity and pro- the agriculture sector, weak diversification, and slow duction mainly due to a lack of access to resources, emigration toward cities as a result of weak job pros- improved technologies, and sources of finance. Nota- pects. A salient feature of Burkina Faso agriculture bly, agricultural production systems rely on extensifi- is that the increase in cereal production is entirely cation and family labor, with limited use of improved explained by the extension of harvested areas, while inputs (Agricultural Productivity and Diversification yields declined by about 3 percent. The agriculture Project PID, 2010; Nutrition Sensitive Agriculture & sector consists mostly of subsistence farming, apart Capacity Building for Small Farmers PID, 2016). from cotton production, which accounts for roughly a The above background highlights the policy attention third of exports. Burkina Faso also has a comparative placed on low agricultural productivity, with issues advantage in mangoes, sesame, and shea nut, but is such as diversification and land extensification also currently exporting at low levels in the value chain for deserving mention. such products (Burkina Faso SCD, 2017; Third Phase food crops grown by the poor and the nonpoor. In Cotton production spans the subregion, par- Côte d’Ivoire, compared to the nonpoor, slightly ticularly in a belt covering the southern parts more of the poor produce maize and fluvial rice of Burkina Faso as well as the northern parts of and slightly fewer produce cacao. In Burkina Faso, Côte d’Ivoire, Togo, and Benin. Interestingly, this slightly more of the poor produce sorghum, mil, cotton belt overlaps with areas of higher agglom- cowpeas, and groundnuts. In Togo, there are more eration that are not close to the capital city of pronounced differences between crops grown by each country. In Benin, cotton is more spread out the poor and the nonpoor, with many more of the throughout the country in terms of proportion of poor producing crops such as cowpeas, yam, sor- individuals planting it, being highest in the north ghum, gombo, soya, and rice. and decreasing going south. Geographical Differences in Agricultural Activity 37 Figure 4.2  Crops Grown by Poor and Nonpoor Côte d’Ivoire Burkina Faso 38 100 100 Poor Poor Nonpoor Nonpoor 81.8 80 80 73.1 60 60 58.5 52.4 53.753.9 56.1 52.1 50.9 46.6 44.8 44.1 40 40 29.4 27.5 29.0 29.2 24.6 % of individuals planting crop % of individuals planting crop 23.6 18.3 19.6 20 16.3 15.8 15.6 15.2 15.0 20 16.4 13.7 14.7 14.4 11.0 12.0 11.2 0 0 Cocoa Yam Maize River Peanut Lower Cassava Coffee Sorghum Mil Maize Cowpea Peanut SeSame Cotton Paddy rice rice rice Togo Benin 100 95.1 100 91.7 Poor Poor Nonpoor 92.8 91.6 Nonpoor 80 80 79.4 78.3 62.3 60 60 55.0 56.5 50.4 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo 46.3 45.0 46.9 40 37.7 40 31.6 34.7 33.6 32.1 27.3 27.3 % of individuals planting crop % of individuals planting crop 25.1 24.5 22.8 24.4 20.4 24.1 20 19.7 18.3 19.2 20 21.0 0 0 Maize Beans, Yam Sorghum Okra Cassava Soya Rice Peppers Cotton Cereals Tubers Fruit and Palm oil Cotton cowpea vegetables Sources: Authors’ calculations using Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. Map 4.2  Cash Crops across Agroecological Zones Sources: Authors' calculations using Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. In agroecological zones where cotton is less of that crop in order to verify whether the crop is prevalent, other cash crops come into play. This in fact a marketable crop. Finally, we considered is true in all countries except in Burkina Faso, which crops are known to be cash crops in that where Zone 4 has the highest proportion of both country. cotton and sesame production, while in Togo, we Côte d’Ivoire shows uniquely diversified pro- see soybean, in Benin palm oil, and in Côte d’Ivoire duction of cash crops across zones: cashew in cacao. In Togo, while Zone 1 dominates in cotton Zones 1 and 2, cotton in Zone 2, and cacao and prevalence, Zones 2, 3, and 4 dominate in soybean some coffee in Zones 3 and 4. About 45 percent production, all of which have over 35 percent of of those engaged in agriculture in Zone 1 and those engaged in agriculture growing the crop. In 42 percent in Zone 2 grow cashew, while fewer Benin, there is an interesting and clear pattern than 5 percent in Zones 3 and 4 grow the crop. whereby cotton production decreases going from Close to 50 percent in Zone 2 grow cotton, while North to South (Zone 1 to Zone 4) while palm oil in all other zones, cotton is non-existent. While production increases. In Zone 4, over 50 percent 70 percent of those in Zones 3 and 4 grow cacao, of those in agriculture produce palm oil, while only only 14 percent in Zone 1 and 7 percent in Zone 2 3 percent grow cotton. In Côte d’Ivoire, cotton is grow the crop. Coffee is grown mostly in Zones 3 grown only in Zone 2, while cacao production is (20 percent) and 4 (13 percent) and by far fewer most prevalent in Zones 3 and 4, with both zones individuals in Zones 1 (5 percent) and 2 (2 percent). having 70 percent of those working in agriculture growing it. The type of food crops grown also varies across agroecological zones, with maize produc- In choosing what other cash crop to observe for tion noted across all countries but to varying each country, we considered three factors. First, degrees. In Côte d’Ivoire, the top crops in Zones we considered the proportion of people growing 1 and 2 are yam and maize, respectively. Maize the crop to ensure that we had a reasonable sam- is less prevalent in Zones 3 and 4, where only 11 ple size that would allow us to observe differences and 12 percent, respectively, produce the crop. across zones. Second, we considered the propor- In Burkina Faso, maize is the most commonly tion of people growing the crop who sold some Geographical Differences in Agricultural Activity 39 Map 4.3  Cash Crops across Côte d’Ivoire’s Agroecological Zones Sources: Authors’ calculations using Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. produced crop in Zones 3 and 4 and is produced in Zone 2 in Togo, and 1,582 kg/ha in Zone 2 in by many in other zones. Fully 20 percent of indi- Benin. This highlights the fact that productivity in viduals in Zone 1 and 36 percent in Zone 2 produce the agriculture sector in Côte d’Ivoire is very differ- maize. Sorghum is produced by many farmers ent from that in the other three countries. across all four zones of Burkina Faso and is one Across agroecological zones within Côte d’Ivoire, of the top two crops in all zones, while mil is par- maize yields are much higher in Zones 3 and 4, ticularly important in Zones 1 and 2. In Togo, maize the same zones where there is prevalent cacao is produced by almost all of those involved in agri- production. Maize yields in Zone 4 are over five culture in all zones, while sorghum production is times higher than maize yields in Zones 1 and 2, more prevalent in Zones 1 and 2 than in Zones 3 even though maize production is most prevalent and 4 (see Appendix D). in Zone 2. Turning to yam yields in Côte d’Ivoire, we note that this crop is not widely produced in Food Crop Yields some zones. However, the yield for yam does not To explore productivity in food crop production, follow the same pattern, being highest for Zones 2 we look at yields, or the ratio of output (measured and 4 and lowest for Zones 1 and 3. in kilograms) to land (measured in hectares). We While over time (2008–2013), cotton, cashew, focus on yields from one food crop, maize, because and rice yields have been increasing, there has this food crop is fairly prevalent across all four been a decrease in yields for important food crops countries and is found in all zones in each country. such as roots and tubers. Moving forward, there This allows for a degree of comparability between needs to be an emphasis on improving food crop yields across zones as well as across countries. yields as these are more important for the north- However, we caution that maize is less prevalent ern areas of the country where poverty rates are in Côte d’Ivoire, where it only largely appears in higher. Beyond this difference in food and cash Zone 3. Data constraints restrict us from being crop yields, there are also large gaps in productiv- able to aggregate different cereal grains, for exam- ity for a given crop across households, suggesting ple, to obtain a measure of cereal yields. scope for productivity gains (Christiaensen and In Côte d’Ivoire, maize yield in its lowest yield zone Lawin, 2017). (1,603 kg/ha in Zone 1) is higher than in highest In Burkina Faso, maize yields are highest in Zone 4, yield zones in any of the other three countries: where conditions are generally favorable. Zone 4 1,234 kg/ha in Zone 4 in Burkina Faso, 1,477 kg/ha 40 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo Map 4.4  Maize Yields across Agroecological Zones Sources: Authors’ calculations using Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. thus sees relatively vibrant agriculture, with strong cotton production. In Togo, Zone 2 is similar to cotton production and relatively high maize yields. Benin in having the highest maize yields in the Despite the fact that conditions are also favorable country. However, unlike Benin, Togo has higher in Zone 3, maize yields are about half in Zone 3 maize yields in Zone 4 than in Zone 3. (683 kg/ha) of those in Zone 4, suggesting large One emerging pattern is that food crop yields variation in maize yields in Burkina Faso. Note that are highest in those zones where key cash crops yields for sorghum do not follow the same pattern, are prevalent. Cash crop production can con- being highest in Zone 2 and mostly similar across tribute to food crop productivity through various all the other zones. pathways. First, cash crop income can help house- In Togo, Zone 3 has the lowest maize yields (817 kg/ holds overcome credit constraints on purchasing ha), while Zones 2 and 4 have the highest. Note fertilizer and other inputs. Second, participation that the disparity in maize yields between low-yield in a resource providing scheme provides access Zone 3 and high-yield Zone 4 is also quite large. to inputs via the marketing firm, and these inputs can in part be used for food crops. Third, cash crop In Benin, differences in productivity across zones income can allow households to make investments are relatively low compared to the other coun- in tractors or draught animals, thereby increasing tries, even though this could be at least partially a food crop productivity. Fourth, technical training reflection of the fact that the Benin data on yields provided by the marketing firm can increase food do not come from household surveys but rather crop productivity (Schneider and Gugerty, 2011; from administrative data at the commune level Strasberg et al., 1999; and Minten, Randrianari- from the Ministry of Agriculture. Thus, there is son, and Swinnen, 2009). In Mali, for example, less comparability between Benin and the three farmers take the credit and fertilizer extended to other countries in the subregion. Maize yields are them by the cotton company and use it on their lowest in Zone 4 (1,074 kg/ha), where agroecologi- food crops (Thériault et al., 2015). The spillover cal conditions should be favorable but also where effects of cash crop production or commercializa- there is no cotton production. Maize yields are tion of agriculture on food crop productivity has highest in Zone 2 (1,582 kg/ha), where there is Geographical Differences in Agricultural Activity 41 been found in other settings, such as for cotton in comparison. Due to the small sample size and low Mozambique and coffee and sugarcane in Kenya prevalence of cash crop production (soya or cot- (Govereh, Jayne, and Nyoro, 1999), cotton in Zim- ton), Togo is also eliminated from this analysis. It babwe (Govereh and Jayne, 2003), and coffee in will be difficult to make any meaningful conjec- Kenya (Strasberg et al., 1999). Positive spillovers tures about differences in cash crop production have also been found at both the household and across agroecological zones in Togo. the regional level in Zimbabwe (Govereh and Jayne, In Côte d’Ivoire, there are no large differences 2003). However, at least in Kenya, the effect of across agroecological zones as regards cash crop cash crop production on food crop yields varies by productivity where cacao is prevalent. Zone 2 has type of cash crop and geographical region, sug- lower yields of cacao (166 kg/ha), and this is also gesting that policies on agriculture commercializa- the zone where cacao is less prevalent. In Zone 2, tion should carefully consider the context. where cotton is prevalent, average cotton yields (956 kg/ha) are similar to those in Burkina Faso Cash Crop Yields and Benin. In Burkina Faso, there are also no differ- We find little variation in cash crop yields across ences in cotton yields across agroecological zones zones in Côte d’Ivoire and Burkina Faso. How- despite the fact that cotton is most prevalent in ever, there is considerable variation in cotton Zone 4. yields in Benin, with higher yields in Zone 1. We In Benin, there is greater variation in cotton yields look at crop yields for the most prevalent cash crop across zones, with Zone 1 having the highest yields in each country in order to compare productivity in (940 kg/ha) and Zone 4 having half the cotton cash crop production across agro­ ecological zones yields of Zone 1. Of the top two cotton produc- within countries. We focus on cacao yields in Côte ing zones in Benin, Zone 1 has higher yields than d’Ivoire and on cotton yields in Benin and Burkina Zone 2. Zone 1 in Benin has similar cotton yields to Faso. However, cotton is not grown in Zone 1 in Burkina Faso’s (close to 1,000 kg/ha), while Zone 2 Burkina Faso and is therefore eliminated from this in Benin has lower crop yields. Map 4.5  Cash Crop Yields across Agroecological Zones Sources: Authors’ calculations using Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. 42 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo Assets, Inputs, and Output to irrigation, and use of fertilizer and pesticides increase the likelihood of commercializing agri- Markets culture produce (Christiaensen and Lawin, 2017). Enhancing agricultural productivity is central The variation in yields we observe across agro- to growth and poverty reduction. While agro­ ecological zones lends itself to the idea that some ecological conditions matter for productivity, areas have lower yields because they have fun- other factors, such as market access, are also damentally poorer growing conditions and are crucial. Various other factors must be in place to subject to poorer soil quality, vegetation, and build productivity and ensure that it translates rainfall. However, this is not necessarily the only into development and poverty reduction (World reason for differences in yields. In fact, we find Bank, 2008). that some poorly endowed zones have reason- able maize yields, such as Zone 1 in Burkina Faso, One set of key ingredients involves increasing and that some better endowed zones have poorer assets such as land, water, and human capital. maize yields, such as Zone 4 in Benin. Prior work The lack of assets is greatest in Sub-Saharan also suggests that cross-country differences in Africa, where farm size in densely populated areas agricultural yields are more likely explained by eco- is falling and investment in irrigation is negligible. nomic decisions than by differences in agroeco- Land markets can raise productivity, help house- logical endowments (Adampoulos and Restuccia, holds diversify income, and facilitate exit from 2017). Instead, these economic decisions are influ- agriculture, even if insecure property rights and enced by institutions, constraints, and policies. poor contract enforcement restrict land markets. Although access to water and irrigation is a major determinant of productivity and stability of yields, Inputs to Production in Sub-Saharan Africa, only 4 percent of the area We now look at three variables regarding inputs under production is irrigated (World Bank, 2008). to production across zones: spending on fertilizer, Another set of key ingredients relates to improv- use of pesticide, and use of irrigation. Due to data ing access to both input and output markets. constraints, the use of pesticide is not recorded Market failures in Sub-Saharan Africa continue in Benin, and the use of irrigation is not recorded to be pervasive in Sub-Saharan Africa because of in Burkina Faso and Togo. high transaction costs, risks, and limited econo- In Côte d’Ivoire, the use of pesticide and aver- mies of scale. Low fertilizer use remains one of the age spending on fertilizer both follow the same major constraints to agricultural productivity in pattern across agroecological zones, where it Sub-Saharan Africa. However, improved produc- is highest in the cotton producing Zone 2. Input tivity produces meager benefits if smallholders use is substantially larger in Zone 2, where cotton cannot sell their produce. Improving access to is produced, than in Zones 3 and 4, where cacao food staples markets and allowing smallholders to is widely produced, and it is lowest in Zone 1. For participate in the production of traditional exports example, the average amount spent on fertilizer can promote faster growth and benefit the poor is about US$620 (at 2011 PPP) in Zones 2 and 4, (World Bank, 2008). with the next highest spending on fertilizer reach- In this section, we map access to assets, inputs, ing only US$100 on average. This difference in the and output markets across agroecological zones use of inputs could also explain the observation in the subregion. Specifically, we take stock of the that the highest yam yields are found in Zone 2 use of fertilizer and phytosanitary products, land and the lowest yam yields in Zone 1. However, the use, the quality of tenure security, and sales of vast difference in fertilizer spending is more likely agricultural produce. Identifying which areas are due to cotton production in Zone 2. weak in terms of access to assets, inputs, and In Burkina Faso, there are equally pronounced output markets is more useful from a policy per- differences across agroecological zones in the spective than learning that various agroecological use of pesticide and average spending on fertil- zones have lower productivity simply because they izer, where it is highest in the cotton producing are naturally disadvantaged. Moreover, to cite the Zone 4 and lowest or even negligible in Zone 1. For case of Côte d’Ivoire, land tenure security, access example, in Zone 4, average spending on fertilizer Geographical Differences in Agricultural Activity 43 Map 4.6  Use of Inputs and Farm Land across Agroecological Zones (a) Fertilizer spending (b) Pesticide use     (c) Irrigation use (d) Farm land size     Sources: Authors’ calculations using Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. is about US$580 (at 2011 PPP), while in Zone 3, spending in Zones 1 and 2. In Togo, particularly which also has some cotton production, spending in Zone 1, US$80 (at 2011 PPP) is spent on fertil- on fertilizer is only US$130. There are also large izer on average, and US$50 in Zone 2. Fertilizer differences in pesticide use across agroecological spending is lowest to negligible in Zone 4. Thus, in zones. These observed differences in input use Togo, the pattern whereby maize yields are high- could explain the high maize yields in Zone 4. How- est in Zones 2 and 4 does not match the pat- ever, it does not necessarily explain why maize tern for fertilizer spending. In Benin, the highest yields are lowest in Zone 3 because this zone has spending on fertilizer is in Zones 1 (US$110) and 2 slightly more input use than Zones 1 and 2. (US$140) where cotton is produced, and spending is only about half that in Zone 4 (US$55). Compar- Compared to Côte d’Ivoire and Burkina Faso, Togo ing Zones 3 and 4, fertilizer spending is higher in has low pesticide use across all zones. However, Zone 3 than in Zone 4, which could partially explain similar to both Côte d’Ivoire and Burkina Faso, the low maize and cotton yields in Zone 4. fertilizer spending in Togo is highest in cotton producing zones. Benin also has high fertilizer 44 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo Land Use Burkina Faso. Zones with favorable agroecological conditions such as Zone 4 in Benin and Togo tend Not only do the cotton producing zones have to use the least amounts of fertilizer. However, we high input use, farm land areas are also larger should note that zones with the poorest agroeco- on average in these zones. We looked at average logical conditions, such as Zone 1 in Burkina Faso, farm size among those who engage in agriculture. also spend the least on fertilizer. In Côte d’Ivoire, not only is irrigation, fertilizer, and pesticide use highest in the cotton produc- ing Zone 2, average farm land size is also highest Land Tenure Security (8.4 hectares), or double that in the other zones. In rural West Africa, the allocation and enforce- The same is true for Burkina Faso, where farm size ment of land rights mostly operates through a is largest for cotton producing Zone 4 (6.3 hect- diverse and overlapping set of customary arrange- ares). Interestingly, farm size in Zone 3, which ments at the village or local level. Increasing pres- has some cotton production, is almost half that sure on natural resources and the absence of in Zone 4. In Benin and Togo, Zone 4, which has written documentation regarding land use have favorable agroecological conditions, has smaller given rise to land conflicts over inheritance as well farm sizes on average. In Benin, farm size is high- as disputes between villages, farmers, and pasto- est in Zones 1 and 2 (about 5 hectares), and in ralists. This lack of formal land rights may lead to Togo, farm size is highest in Zones 1 and 3 (also underinvestment and suboptimal yields. In theory, about 5 hectares). the codification of private property rights within Data on irrigation are available only for Côte d’Ivoire an effective legal framework should increase agri- and Benin. Use of irrigation in Benin is quite low, cultural investment and productivity. In this sense, about only one-fifth of such use in Côte d’Ivoire, land reform is a critical policy action in the sub­ where irrigation is lowest in Zone 1 and highest region, as it is elsewhere. in Zone 2. In Benin, irrigation is highest in Zone 1. We first look at reported ownership of land across The general pattern emerges whereby cotton agroecological zones. Across all countries, there growing areas use more pesticide, spend much is relatively high reporting of owning at least one larger amounts on fertilizer, and have larger plot, except in Benin. In Benin, while 76 percent of farms. We see this in Zone 2 in Côte d’Ivoire individuals own at least one plot in Zone 4, fewer (which also uses more irrigation) and in Zone 4 in than 60 percent own at least one plot in all other Box 4.2  Background on land reform in Benin, Côte d’Ivoire, and Burkina Faso We first provide a brief background on land reform in Almost all farmland is owned and transferred accord- some of the countries in the subregion before map- ing to customary law. Because customary land tenure ping out land tenure security across agroecological systems are not well defined or consistently applied, zones. their use has led to conflict. The 1998 land reform legislation, which aimed at moving from customary In Benin, a Rural Land Use Plan (Plan Foncier Rural, land tenure to more modern arrangements, has been PFR) has been in operation since 1993 and has been slow in its implementation (Côte d’Ivoire SCD, 2015; recognized by the 2007 Rural Land Act. It is consid- Agricultural Sector Support Project, PAD, 2013). ered different from more standard land formalization programs in two respects. First, it recognizes that In Burkina Faso, despite land reform carried out by existing customary arrangements provide legitimate the government through a consultative process over claims to property that can be formalized. Second, it the last few years and the consequent adoption of sets up a decentralized procedure for the establish- various legislation, its enactment has been slow. ment of formal property rights. Recent work sug- Only a small fraction of all land has been registered gests that the demarcation process of PFRs lead to formally. Traditional laws for land management and long-term investments in agriculture (Goldstein et al., community-based ownership continue to hamper 2015). land transactions as their relationships with modern laws have not yet been fully clarified (Burkina Faso In Côte d’Ivoire, insecurity of land tenure has been one SCD, 2017). of the root causes of conflict and constrains invest- ment in agricultural development and agro-business. Geographical Differences in Agricultural Activity 45 Map 4.7  Land Tenure Security across Agroecological Zones Sources: Authors’ calculations using Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. zones. In Burkina Faso and Côte d’Ivoire, compared Access to Output Markets to other agroecological zones within each coun- Achieving reductions in poverty via enhanced pro- try, there is only slightly less ownership of land in ductivity and improved input and land markets is the cotton producing zones, particularly Zone 4 unlikely to be achieved if households are unable in Burkina Faso (70 percent) and Zone 2 in Côte to access markets to sell their produce. The lack d’Ivoire (74 percent). of access to markets inhibits the poor from par- Among those who own land, we look at the securi- ticipating in the benefits of improved agricultural tization of land tenure across agroecological zones. productivity (Schneider and Gugerty, 2011). In this There are only slight differences in the definition section, we map out the sales of crops across of securitization across countries. In Côte d’Ivoire agroecological zones. Due to data constraints, and Togo, weak securitization is defined as having data on the sale of crops are not available in Benin. no form of documentation for at least one plot In Côte d’Ivoire, there is not much variation owned. In Burkina Faso and Benin, weak securiti- across zones in the proportion of individuals zation is defined as having no form of documen- who sell some of their crop, even though the tation or verbal agreement for at least one plot value of sales is substantially higher in Zone 2. In owned. Thus, our definition of “securitization” is Zones 1 and 2, about 73 percent of individuals sell less stringent for Burkina Faso and Benin (which some of their crop compared to about 82 percent includes verbal agreements for tenure security) in Zone 4. However, among those who sell some of than for Côte d’Ivoire and Togo. their crop, the value of sales varies widely across Across countries, there seems to be weaker zones. Zone 2 has the highest average value of securitization of land in higher value agricul- crop sales (about US$3,800 at 2011 PPP), followed tural areas, particularly in cotton producing by Zone 4 (US$2,700). The average value of crop zones. Among those who own some land, there sales in Zone 1 is less than one-third of the average is better securitization of ownership in Zone 4 in value of crop sales in Zone 2, and that in Zone 3 both Benin and Togo. In Côte d’Ivoire, compared is only about half of the value of sales in Zone 2. to Zones 1 and 2, there are more individuals with In Burkina Faso, very few sell their crop in better securitization of land in Zones 3 and 4. In Zone 1, while many more do so in Zone 4, where Burkina Faso, where ownership is low, securitiza- the value of sales is also substantially higher. tion of land is also lower in Zone 4. There is a great deal of variation in access to crop markets in Burkina Faso, where only 30 percent 46 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo Map 4.8  Sale of Agricultural Produce (a) Sell some of their crop (b) Revenue from crop sales   Sources: Authors’ calculations using Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. of individuals in Zone 1 sell some of their crop, Burkina Faso, there is greater market exchange of whereas 73 percent do so in Zone 4. Among those crops in Togo. In Togo, the zones with the highest that sell some of their crop, the average value of proportion of individuals who sell some of their crop sales is about four times higher in Zone 4 crop are those zones with lower maize yields. This (US$2,200 at 2011 PPP) than in Zones 1 and 2. is unsurprising as maize is primarily a food crop. Moreover, the average value of crop sales is also Thus, these zones are likely to be selling crops over three times higher in Zone 4 than in Zone 3. other than maize. Among those who sell some of their crop, the average value of crop sales is In Togo, both the proportion of individuals who much higher in Zone 1 (US$1,500 at 2001 PPP) sell some of their crop and the value of crop sales and 3 (US$1,100) than in Zones 2 (US$600) and among those who do so is highest in Zones 1 4 (US$700). and 3. In Zone 1, 86 percent sell some of their crop, while in Zone 3, 79 percent do so. Compared to Geographical Differences in Agricultural Activity 47 Chapter 5 Putting It All Together Having graphically established a two-way relation- How does this play out in the case of Benin, Burkina ship between welfare and each of the three build- Faso, Côte d’Ivoire, and Togo? The answers are: ing blocks of the economic geography literature 1. A region’s income levels and its agricultural discussed in Chapter 3, and carried out a separate productivity are closely related in the four process for agricultural productivity in Chapter 4, countries studied; we now put all the factors together. Specifically, in this chapter, we aim to quantify the relationship 2. Except for being coastal or landlocked, geo- between welfare and agricultural productivity. graphical covariates do not appear to be Moreover, we seek to analyze correlates of varia- directly associated with a region’s per capita tion in both welfare and agricultural productivity income. However, the relationship between across space. To do so, we define three sets of geography and welfare is mediated by explanatory variables corresponding to the three agglomeration economies and market access; building blocks of the economic geography: natural 3. As expected, there is a strong link between endowment, agglomeration, and market access. agroecological characteristics and yields per Our dependent variables are welfare and agricul- hectare; and tural productivity. 4. What is notable is the persistence of the cor- Welfare levels are measured by average house- relation between geography and crop yields hold consumption in each administrative unit.18 regardless of whether population density, Hereafter, we use consumption and income inter- market access, or farm inputs are controlled changeably. While different cash crops and food for. This pattern should be taken into consid- crops are harvested across space (see Figure C.3 eration when planning development strate- in Appendix C for a complete list of crops), data gies in the agriculture sector. on maize yields (in kg/ha) are available across all four countries and in most administrative units and are therefore used here as a proxy for agri- cultural productivity. In this paper, we provide Relationship between only a static picture of the relationship between Welfare and Agricultural welfare, agricultural productivity, and the three sets of explanatory variables. The results should Productivity therefore be interpreted as correlates only. We first show that agricultural productivity is There has been an ongoing debate in the economic positively correlated with per capita expendi- geography literature about whether a location’s ture. This rationalizes the geography of agricul- levels of per capita income and other economic tural productivity as a means to better understand dimensions are determined by geographical and the geography of poverty. For Benin, Burkina Faso, ecological variables. Many researchers have pro- Cote d’Ivoire, and Togo, we find that on average, a vided evidence supporting the view that such links 10 percent increase in mean maize yields is associ- are strong (see, for example, Gallup et al., 1999; ated with a 1.7 percent increase in mean per capita Sachs, 2000; Gallup and Sachs, 2001; Sachs and expenditure (Figure 5.1). Malaney, 2002), while others have argued that the The relationship between higher productivity role of geography in explaining spatial patterns of and lower poverty in the four countries is con- per capita income operates through various direct sistent with what has been observed in other channels (e.g., productivity and trade) or indirect countries. Globally, there is an almost one-to- channels (e.g., choice of political and economic one relationship between crop yields and poverty institutions) with little direct effect of geography alleviation such that a 1 percent increase in agri- on income (see, for example, Acemoglu, Johnson, cultural productivity is correlated with a 0.9 per- and Robinson, 2001; Easterly and Levine, 2002; cent reduction in poverty (Irz et al., 2001). In India, Rodrik, Subramanian, and Trebbi, 2004). the poor have gained in both absolute and rela- tive terms from increased farm yields (Datt and Ravallion, 1998). In Madagascar, higher yields are linked to improvements in food security (Minten 18 Administrative units are communes in Benin, provinces in Burkina and Barrett, 2008). In Ethiopia, more complex Faso, departments in Côte d’Ivoire, and prefectures in Togo. Putting It All Together 49 Figure 5.1  Correlation between Poverty and Agricultural Productivity 1 log (consumption per capita) .5 0 –5 −1 −4 −2 0 2 4 log (maize yield) b = 0.17*** (including country fixed effects) Sources: Authors’ calculations based on Benin EMICOV 2015, Burkina Faso EMC 2014, Côte d’Ivoire ENV 2015, and Togo QUIBB 2015. measures of agricultural productivity also corre- We start with natural endowment covariates con- spond to reductions in poverty (Abro, Alemu, and sisting of six continuous agroecological variables Hanjra, 2014). There are multiple pathways from that arguably affect a region’s agricultural produc- farm productivity to poverty reduction. Beyond tivity and therefore income. These are tempera- the potential direct effects on farmers, there are ture, precipitation, soil quality, elevation, latitude, also indirect effects such as nonfarm job creation and ruggedness.19 While high elevation locations as well as linkages to the rest of the economy. often have low market access due to poor trans- Improvements in productivity can lower food portation, once ruggedness and coastal locations prices (which benefits poor net buyers of food) are controlled for, it is best interpreted as an agro- and raise the wages of unskilled workers (which ecological variable. We thus introduce a seventh benefits poor unskilled workers). However, food variable, namely a coastal dummy, which takes a prices and wages may change slowly over time value of 1 if the administrative unit has a coast such that the longer term effects of agricultural and zero otherwise. We then add an agglomera- productivity on poverty reduction may outweigh tion economies variable measured as the log of the short-term effects (Schneider and Gugerty, the number of people per square kilometer. For our 2011; De Janvry and Sadoulet, 2010; Minten and regressions on agricultural productivity, we also Barrett, 2008; Datt and Ravallion, 1998). include the share of employed population in the agriculture sector. Finally, our covariate for market access is the log of market access index described Three Sets of Explanatory in Chapter 2. Table 5.1 reports the statistics of the Variables for Three Building variables used. Blocks The next question is: Among the three building blocks of economic geography, which factors are 19 Two additional agroecological variables—length of growing period significantly associated with a location’s welfare and slope—are strongly correlated with latitude and precipitation and agricultural productivity? (> 0.9) and were therefore are excluded from our regressions. 50 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo Table 5.1  Statistical Summary Number of Standard Observations Mean Deviation Dependent variables Consumption per capita (2011 PPP) 265 893 355 Maize yields (kg/hectare) 237 1,918 2,823 Natural endowment covariates Temperature (celsius) 263 26.891 0.927 Precipitation (mm/year) 261 1,171 268 Soil 261 4.083 2.584 Latitude 261 8.628 2.314 Elevation (m) 261 229 119 Ruggedness 263 0.202 0.185 Coastal dummy 265 0.064 0.245 Agglomeration covariate Population density (people/km2) 265 244 893 Share of population in agriculture 264 0.605 0.272 Market access covariate Log (market access index) 265 9.825 2.036 Note: Observations are recorded at administrative unit levels: 77 communes in Benin, 45 provinces in Burkina Faso, 107 departments in Cote d’Ivoire, and 36 prefectures in Togo. Correlates of Welfare than the South. Keeping other factors constant, cooler temperatures and less rugged terrain are Table 5.2 presents the results of the multivariate also associated with higher income. However, rain- regressions in which we relate an administrative fall levels, soil quality, and elevation do not seem unit per capita consumption level to the three to be directly linked to income. sets of explanatory variables described above. To The second observation is that being on the assess spatial differences within a country, we coast signals wealth (Column 2). On average, per include country fixed effects in all of our regres- capita consumption in a coastal area is 55 per- sions. Column 1 looks at the relationship between cent higher than in a landlocked location. After consumption and six agroecological characteris- controlling for whether or not a location is on the tics. We are particularly interested in the relative coast, the income pattern from North to South effect of each characteristic on welfare. Column 2 still holds. A one degree increase in latitude is cor- adds the seventh variable of interest, a dummy related with a 10.2 percent decrease in per capita variable that takes the value 1 if the administra- consumption. Ruggedness remains a key variable tive unit has a coast and zero otherwise. Column 3 with a significant link to a location’s income, while introduces the agglomeration economies variable the remaining four natural endowment covariates of population density. Finally, Column 4 adds the (temperature, rainfall levels, soil quality, and eleva- market access covariate. tion) do not. The first observation is that the closer a loca- Interestingly, once population density is taken tion to the equator, the higher its income, with into account, none of the agroecological char- the coefficient of the latitude variable being nega- acteristics, except the coastal dummy, is sig- tive and significant (Column 1). More precisely, an nificant (Column 3). In other words, a location’s increase of one degree of latitude is correlated wealth is solely related to its level of agglomera- with a 9.4 percent drop in per capita consump- tion and its situation near the sea. Between two tion. This confirms our story described in Chap- locations with the exact same population density, ter 3, namely that the North is generally poorer the one on the coast is 17.5 percent richer than Putting It All Together 51 Table 5.2  Factors Associated with Spatial Differences in Poverty: Coastal Location, Population Density, and Market Access Regression Results (1) (2) (3) (4) Dependent variable Log (consumption per capita) Natural endowment covariates Temperature (Celsius) –0.129* –0.079 –0.032 –0.028   (0.075) (0.072) (0.066) (0.065) Precipitation (mm/month) 0.000 –0.000 0.000 –0.000 (0.000) (0.000) (0.000) (0.000) Soil quality (per mille) –0.002 –0.013 –0.010 –0.007 (0.020) (0.018) (0.012) (0.012) Latitude (degrees) –0.094* –0.102** –0.039 –0.047 (0.051) (0.043) (0.035) (0.034) Elevation (m) –0.001 0.001 0.000 0.000 (0.001) (0.001) (0.000) (0.000) Ruggedness (in 100 m) –0.451*** –0.301* –0.149 –0.136 (0.169) (0.157) (0.134) (0.128) Coastal dummy 0.545*** 0.175** 0.210*** (0.088) (0.071) (0.068) Agglomeration covariate Log (population density) 0.157*** 0.092*** (0.024) (0.032) Market access covariate Log (market access index) 0.046*** (0.017) Country fixed effects Yes Yes Yes Yes Number of observations 261 261 261 261 Adjusted R squared 0.525 0.626 0.711 0.724 Note: Standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1. one located inland. Similarly, holding other vari- What these results suggest is that with the ables constant, a 10 percent increase in population exception of being landlocked, the relationship density is associated with a 1.6 percent increase between geography and welfare is mediated by in income. population density and market access. In other words, areas good for growth will either attract A key message from Column 4, which includes people or experience stronger population growth all three sets of explanatory variables (natural and at the same time receive investments in infra- endowment, agglomeration, and market access) structure. Thus, when controlling for population is that all three sets matter for a region’s income density and market access, we no longer detect in Benin, Burkina Faso, Côte d’Ivoire, and Togo. any relationship between welfare and tempera- However, a region’s position in relation to the sea ture, latitude, or ruggedness. is the only geographic characteristic associated with that region’s wealth. With the same levels As shown in Chapter 3, agglomeration and mar- of agglomeration and market access, a coastal ket access often go hand in hand (i.e., densely location’s income is 21 percent higher than that populated areas enjoy better market access and of a landlocked one. vice versa). Therefore, after controlling for market 52 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo access levels, the relationship between agglom- are temperature, latitude, elevation, and location eration and income declines in magnitude, albeit in coastal areas. Similar to the relationship with remaining significant (Columns 3 and 4). What income, temperature is also negatively correlated Column 4 implies is that keeping all variables with agricultural productivity: on average, one constant, including market access, a 10 percent additional degree Celsius is associated with an increase in population density correlates with a 0.9 87 percent decrease in agricultural productivity. percent increase in wealth compared to 1.6 per- Similarly, land located at high elevations has lower cent in Column 3, where market access is not crop yields, with a 100-meter higher location cor- accounted for. In addition, a 10 percent improve- relating with a 70 percent reduction in production. ment in market access is associated with a 0.5 Notably, being on the coast signals not only percent increase in income. wealth but also high agricultural production. While the results suggest that human activ- We find that a coastal region is associated with ity does respond to geography in the sense that a more than 100 percent increase in yields. How- people move or build roads where the environment ever, controlling for a region’s location (whether is more favorable, we find that the benefits of landlocked or coastal), the further a region from being along the coast are not arbitraged away by the equator, the higher its agricultural yield, while migration or increased market access. This is an one degree of latitude toward the North corre- interesting finding in that it reveals the untapped lates with a 44.6 percent increase in agricultural potential for economic development provided by productivity. This relationship is opposite to that access to international trade for the three coastal between latitude and income presented in the pre- countries (Benin, Côte d’Ivoire, and Togo). vious section (Table 5.2, Column 2), suggesting that economic activity near the equator (i.e., the South) must expand beyond the agriculture sector to make up for the decline in crop yields. This pat- Correlates of Agricultural tern is consistent with the distribution of sectoral Productivity employment analyzed in Chapter 4, which showed that the concentration of the employed popula- In addition to the three sets of covariates of tion in the industry and service sectors increases natural endowment, agglomeration, and mar- toward the South (Table 5.3). ket access, this section explores the agriculture In contrast to NEG literature suggesting that module found in the household consumption sur- agglomeration economies and market access veys in order to select farm inputs that matter can help farmers take advantage of better prices for agricultural productivity. The challenge is to and a wider selection of agricultural labor, farm identify a set of farm input variables that are not inputs and technology, and better markets for only common across surveys in Benin, Burkina harvested crops, we do not find such a link in Faso, Côte d’Ivoire, and Togo but also available the four countries under study. The correlations in most administrative units. The three variables between agglomeration covariates, market access that meet such requirements are: spending on index, and agricultural productivity are not sig- fertilizer, land size, and share of farmers having nificant (Columns 2 and 3). This finding suggests weak land tenure security. that we may in fact be dealing with two types of In Table 5.3, column 1 presents the correlation agriculture: a subsistence agriculture, whereby between agricultural productivity and agroecologi- most crops are cultivated for home consump- cal characteristics, including the coastal dummy. tion and where investments are less sensitive to Column 2 adds agglomeration variables consisting market access, and a commercial agriculture con- of population density and share of the employed centrated along the coastline and benefiting from population working in the agriculture sector. Col- higher investments in inputs (such as fertilizer). umn 3 introduces market access, and Column 4 When we introduce farm input covariates into adds three farm input variables. the model (Column 4), only the value of expen- Our results in Column 1 show that 4 out of the diture on fertilizer shows a strong correlation 7 agroecological characteristics under study are with crop yields. A 10 percent increase in fertilizer important to agricultural productivity. These spending is associated with a 1.1 percent increase Putting It All Together 53 Table 5.3  Role of Geographical Differences in Agricultural Productivity, Natural Endowments (temperature, latitude, elevation, coastal location), and Spending on Fertilizer Regression Results (1) (2) (3) (4) Dependent variable Log (maize yield) Natural endowment covariates Temperature (Celsius) –0.873*** –0.870*** –0.859*** –0.961*** (0.266) (0.269) (0.265) (0.317) Precipitation (mm/month) 0.001 0.001 0.001 –0.000 (0.001) (0.001) (0.001) (0.001) Soil quality (per mille) –0.048 –0.051 –0.049 –0.035 (0.044) (0.045) (0.045) (0.047) Latitude (degrees) 0.446*** 0.415*** 0.401*** 0.351** (0.120) (0.131) (0.132) (0.146) Elevation (m) –0.007*** –0.007*** –0.007*** –0.007*** (0.002) (0.002) (0.002) (0.002) Ruggedness (in 100 m) –0.576 –0.522 –0.505 –0.237 (0.487) (0.501) (0.496) (0.503) Coastal dummy 1.081*** 1.105*** 1.166*** 1.148*** (0.248) (0.306) (0.311) (0.301) Agglomeration covariate Log (population density) –0.085 –0.170 –0.097 (0.093) (0.111) (0.113) Share of population in agriculture –0.501 –0.358 –0.456 (0.373) (0.348) (0.339) Market access covariate Log (market access index) 0.074 0.040 (0.045) (0.045) Agricultural input covariates Log (fertilizer spending) 0.107*** (0.038) Log (land size) 0.115 (0.138) Land title dummy –0.188 (0.309) Country fixed effects Yes Yes Yes Yes Number of observations 233 232 232 218 Adjusted R squared 0.498 0.503 0.507 0.527 Note: Standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1. 54 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo in agricultural productivity. However, land owner- natural endowment and crop yields. The sign and ship regime and land size do not appear to play a magnitude of the coefficients for temperature, significant role in yields above and beyond their latitude, elevation, and the coastal dummy barely potential influence through fertilizer use. change regardless of whether or not variables for agglomeration, market access, or farm inputs are An important pattern that emerges from the taken into account. four regressions is the persistent link between Putting It All Together 55 Chapter 6 Policy Recommendations and Further Studies Over the past five years, GDP in Benin, Burkina there is a strong need for tailoring inputs and Faso, Côte d’Ivoire, and Togo has grown at the R&D to specific agroecological zones. impressive rate of approximately 5 percent annu- 3. Fiscal transfers: There exist geographical ally. However, such economic growth has not yet pockets of poverty where the costs of reach- translated into high levels of prosperity. Poverty ing the poor are very high. These areas are rates remain relatively high, even by SSA stan- characterized by a combination of high pov- dards. Moreover, spatial disparities in welfare and erty rates and low poverty density. Another poverty are evident throughout all four countries, set of lagging areas with little prospect of with a typical household in a leading area con- growth consists of those with unfavorable suming as much as 7 times more than a similar agroecological characteristics and limited household in a lagging area. Within countries, opportunities to diversify into nonagricultural there is also strikingly wide variation in poverty sectors. Our findings imply that some lagging rates across administrative units. areas may not be able to improve their wel- As illustrated in the World Development Report— fare after all. This may call for pro-poor fiscal Reshaping Economic Geography (World Bank, transfers through a system of interregional 2009), within-country disparities can be a poten- transfers to ensure equity across the leading tial source of tensions between lagging and lead- and lagging regions. Since there are limited ing regions and may affect the country’s overall job opportunities in low-density and lagging growth and political stability. How can our find- areas, it is important that local governments ings help policy makers reduce geographical dif- use the fiscal transfers received to invest in ferences in welfare while boosting growth? Based portable assets (health and education) for its on our analysis, we propose four broad policy citizens so that they can join a healthy and recommendations: educated workforce if they choose to migrate. 4. Safety net programs: Not all poor people, 1. Urbanization: We find that many of the lead- especially the vulnerable, can benefit from the ing areas have not yet maximized the ben- policies proposed above. The need to main- efits of agglomeration economies, especially tain strong safety net programs targeting the in Burkina Faso and Côte d’Ivoire. Based on poor and the vulnerable remains strong. New NEG literature, there is scope for increasing technologies such as e-vouchers and mobile the concentration of economic activities and transfers make it possible for such programs labor in these areas to further take advantage to reach targeted beneficiaries in low-density of economies of scale and boost economic areas in a cost-effective way. Moreover, safety development. However, it is important to con- net programs should be part of an overarch- sider complementary policies to urbanization, ing poverty reduction strategy consisting of including removing barriers to labor mobility interacting with and working alongside urban so that people can migrate to leading areas policy, agricultural productivity boosting pro- where labor demand and productivity are grams, and other policies aimed at eradicating higher, and investing in urban infrastructure poverty and reducing vulnerability. and the provision of public services to accom- modate a potentially larger number of users. 2. Increasing agricultural productivity: Not all Urbanization rural families can move to urban locations. For those staying in the agricultural sector in rural We observe a strong link between agglomera- areas, policy makers may consider improving tion economies and income levels in the four their welfare by increasing agricultural pro- countries. However, many of the leading regions ductivity. Potential areas of improvement still have low population density, i.e., fewer than include land tenure, irrigation, use of farm 150 people per square meter, especially in Burkina inputs such as fertilizer, and research and Faso and Côte d’Ivoire. Such low population den- development (R&D). Given that agro­ecological sity makes it difficult for these regions to reap the endowments seem to influence crop yields benefits of agglomeration productivity and thus more than agglomeration and market access, further advance their economic development. Policy Recommendations and Further Studies 57 Arguably, an urban agglomeration economy early and coordinated urban infrastructure and brings many economic benefits. The first and public service investments. A formal market for foremost advantage is a reduction in transpor- urban land offers buyers the legal protection of tation costs for goods as producers are located the government and generates the public good of near their customers. In the 1990s, New York accurate valuation. Not only is this a precondition and London were manufacturing powerhouses for land consolidation, which converts low-density as factories were built in and around these cit- residential use into higher density housing or clus- ies for better access to customers and transpor- ters of new commercial structures, but it is also tation services (Lall, Henderson, and Venables, an incentive for farmers to invest in inputs such 2017). Moreover, the advantage of agglomeration as fertilizer given that the risk of expropriation is economies increases with scale. Rosenthal and now lower thanks to more secured property rights. Strange (2004) show that each doubling in city In addition, the early installation of coordinated size increases productivity by 5 percent, and the urban infrastructure helps shape urban structures elasticity of income with respect to city population and save costs. If postponed until after population is between 3 and 8 percent. settlement, such infrastructure is far more diffi- cult and expensive to install. Another important Urbanization is strongly associated with pro- policy for urbanization is the provision of public ductivity gains through their links to structural goods and services to ensure quality of life for an transformation and industrialization. As a coun- increasing urban population (Lall, Henderson, and try urbanizes, people move from rural to urban Venables, 2017). areas in search of better job opportunities in terms of higher pay or productivity. Thus, a complemen- tary and necessary policy to favor urbanization is the removal of barriers to labor mobility so that people are able to not only physically migrate to Agricultural Productivity urban cities but also to move to other economic It is important to improve welfare in rural areas. sectors that offer better returns. Not all rural families can move to urban cities, for A highly dense location (in population or firms) many reasons, including limited absorptive capac- also brings down the costs of certain public ity of denser areas, government policies preventing investments such as infrastructure and basic slum proliferation, poorly defined land ownership, public services. The average costs of such pro- risk aversion, or poor information (Gollin, Kirch- grams are lower when their users are many in berger, and Lagakos, 2016; de Brauw, Mueller, and number and densely grouped together. Input costs Lee, 2014). For those remaining in the agriculture for firms located near each other also decline as sector in rural areas, policy makers may look to firms share infrastructure and suppliers. In addi- improve their productivity, which has been shown tion, thick labor markets allow firms to reduce in our analysis to have a strong link with welfare. search costs and to have access to a larger pool In this report, we illustrate how securing land of potential workers. This reduction in mobility tenure and irrigation are largely untapped in the costs thus allows improved allocation efficiency. subregion. Furthermore, the use of inputs such Close spatial proximity also promotes innovation as pesticide and fertilizer is low, which suggests and knowledge sharing among people and firms. the potential for targeted farmer subsidies and International evidence illustrates how knowledge the need for finance so as to enable access to spillovers are a crucial element in improving the credit for the purchase of inputs and machinery. productivity of successful urban cities. East Asia’s Access to savings and insurance will also be crucial success story (e.g., China, the Republic of Korea, in allowing farmers to mitigate both idiosyncratic Vietnam) shows a strong relationship between and covariate weather shocks. urbanization and economic development (Lall, The World Bank’s World Development Report Henderson, and Venables, 2017). (WDR, 2009) recommends two broad sets of To boost growth, policy makers may consider effective instruments for leveraging agriculture promoting urban planning in some leading for development. The first is to increase access to regions. Policies that support urbanization may assets such as land, water, education, and health. include formalizing land markets and making The second is to make smallholder farming more 58 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo productive and sustainable. These instruments tariffs, and input subsidies), human capital (edu- and goals are naturally intertwined as access to cation and health), infrastructure (extent of roads), assets enhances productivity, and sustainable and the prevention of armed conflicts. Although smallholder farming is a prerequisite to securing research may have long-lasting effects on produc- productivity in the longer term. tivity, that effect may be delayed. Fuglie and Rada find that for every US$1 spent on R&D, US$3– More specifically, WDR (2009) emphasizes US$5 in benefits are generated. Higher returns to several possible instruments such as improv- agricultural research in SSA could be seen through ing price incentives and increasing the quality strengthening the CGIAR system, followed by and quantity of public investment, making out- strengthening national agricultural systems in put and input markets work more efficiently, larger countries. Beyond R&D investments that improving access to financial services and have direct effects on TFP growth, policies that reducing exposure to uninsured risks, enhanc- strengthen the enabling environment are also ing the performance of producer organizations, crucial to raising agricultural productivity in the and promoting innovation through science and region. Economic policies that reduce net tax on technology. In Sub-Saharan Africa, market fail- the agriculture sector (i.e., subsidies) and increase ures in input markets continue to be pervasive, the levels of labor force education are found to and as a result, low fertilizer use is one of the increase agricultural productivity. On the other major constraints on increasing agricultural pro- hand, armed conflict and the spread of Human ductivity in the region. Although providing fertil- Immunodeficiency Virus/Acquired Immune Defi- izer subsidies is a clear strategy, the focus should ciency Syndrome (HIV/AIDS) are barriers to agri- shift toward more sustainable solutions such as cultural productivity. targeted vouchers to enable farmers to purchase inputs and stimulate demand in private markets Given that agroecological endowments seem to as well as providing matching grants to underwrite influence agricultural yields more than agglom- start-up costs of entry into markets for private eration and market access, policies designed to distributors. Which instrument or combination of improve agricultural productivity should be sen- instruments will prove most effective in enhancing sitive to these agroecological differences. R&D productivity will depend on the context, not only of efforts and the available input mix (such as fertil- each region or country but also within each region izer blends and seed varieties) should be localized and for each value chain. Furthermore, innova- to agroecological zones. In Mali, for example, the tions in information and communication technol- same fertilizer mix that is suitable for cotton is ogy (ICT) can be leveraged to make markets work inappropriately used country wide despite the fact better through interventions known collectively that such mix is less effective outside of cotton as “e-agriculture.” For example, both Benin and production. Burkina Faso have seen projects launched to col- lect and disseminate food market prices by short message service (SMS). In Côte d’Ivoire, a mobile phone program called CocoaLink connects farm- Fiscal Transfers ers to agricultural experts who can address ques- Our findings highlight the stylized fact that tions concerning fertilizer application or disease some lagging regions may be falling further and pest control in real time. Thus, e-agriculture behind as a result of the development process. can spread information on market prices and best Over time, economic geography will continue to farming practices. favor economic concentration in leading areas and Fuglie and Rada (2012) compare a set of policies make it more difficult for poor areas to catch up. that may correlate with total factor productiv- A set of locations that are particularly vulnerable ity (TFP) growth across countries in Sub-Saharan to this development path are areas with not only Africa, namely investment in research (through high poverty rates but also low poverty densities. national and international agricultural research Such a combination leads to extremely high cost centers such as Consultative Group for Interna- of investments in infrastructure (e.g., market tional Agricultural Research [CGIAR]), economic access, irrigation) and public services (e.g., elec- policies (commodity price interventions, trade tricity, water). Another possible set of vulnerable Policy Recommendations and Further Studies 59 locations is characterized by unsuitable agroeco- Arguably, safety net programs can play cru- logical characteristics for agricultural production cial roles in development policy (Grosh et al., and a lack of job opportunities in nonagricultural 2008). First, they aim to redistribute income to sectors. Evidence from our analysis points to a the poorest and the most vulnerable, resulting in persistent relationship between natural endow- an immediate impact on poverty and inequality. ments and agricultural productivity regardless Second, safety nets can allow households to take of the levels of agglomeration, market access, or up investments both in terms of human capital farm inputs used. and financial investments with a view to securing their future. Third, such programs help the poor One option is to create opportunities for resi- and vulnerable manage risks. Finally, they can free dents to move to higher return areas should they other sectors from the role of redistribution and choose to do so. Policies supporting this option concentrate on efficiency instead. are described in the section on urbanization above. However, not everyone wishes to move for various However, it may be prohibitively expensive reasons, such as cultural barriers (e.g., ethnicity, for these programs to reach the poor in some language) or risk aversion. areas. As illustrated in our analysis, such areas are characterized by high poverty rates but also This calls for pro-poor fiscal transfers. Lagging low poverty density. A general payment system areas have a low base of economic activity to tax would require distributing agencies such as non- and thus yield low revenues. This budget con- governmental organizations (NGOs), public agen- straint prevents them from providing adequate cies, banks, and retail stores to physically deliver social safety nets for poor residents, who repre- cash, in-kind goods, or vouchers to beneficiaries sent a large share of the population. Moreover, it in sparsely populated areas where transportation limits the ability to fund investments in human costs are high. and physical capital and deliver public services, all of which are likely to have extremely high deliv- Fortunately, innovative and affordable tech- ery costs per user. Achieving equity through fiscal nology such as electronic and mobile vouchers transfers can therefore ensure a level playing field allow such programs to cover low-density areas (World Bank, 2010). efficiently. For example, targeted beneficiaries can receive vouchers automatically on their phone. However, transferring financial resources to They can redeem these vouchers for cash or in- lagging regions alone may not be sufficient. This kind goods at any participating retailers or agency should be accompanied by improving capacity and at any time. This process not only saves delivery accountability on the part of local government. costs but also provides transparency and over- Moreover, since there are limited job opportunities comes payment delays as well as complicated and in low-density and lagging areas, it is important inconvenient redemption systems (World Food that local governments use the fiscal transfers Programme, 2014). received to invest in portable assets (health and education) so that citizens can join a healthy and educated workforce if they choose to migrate (World Bank, 2010). Limitations and Further Studies Safety Net Programs This report aims to provide policy makers with stylized facts about spatial differences in wel- There remains an urgent need to maintain fare, poverty, and agricultural activity in Benin, strong safety net programs as the proposed pol- Burkina Faso, Côte d’Ivoire, and Togo. Our main icy above may not reach all those in need. Ultra- challenge was to obtain welfare and agricultural poor people may not have the financial means or data that are comparable over time and across adequate information to migrate to urban areas. countries. Given the low frequency of household Poor women may not have access to farming or consumption surveys and changes in instruments other job opportunities. Poor children may not be and methodologies between surveys, we can only able to attend school due to home–school dis- observe a static pattern of economic geography tance or family financial constraints. in the subregion. In making comparisons across 60 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo countries, we were limited by a small subset of Harmonization Project have the potential to variables available in all four countries as well as make this research feasible. in most administrative units within each country. 2. Labor mobility and migration: We have high- This clearly narrowed the scope of our analysis. lighted the importance of human mobility in For example, only maize yields could be used as a economic development and poverty reduction. proxy for agricultural productivity, and farm inputs Like other reallocation mechanisms, labor were limited to fertilizer, land tenure, and land size. mobility allows workers to migrate to geo- As pointed out in Chapter 2, another important graphic locations or economic sectors where caveat concerns the risk of imprecise estimates returns are higher, thus boosting productivity and lack of representativeness when presenting and economic growth. Labor mobility can also the data at lower administrative unit levels than empower traditionally disadvantaged groups, the surveys were designed for. Finally, our calcula- especially women (World Bank, 2010). How- tion of a market access index may underestimate ever, there is little information on migration coastal areas as well as locations along a country’s in the four countries or on its role in economic borders. development. A detailed study of labor mobil- On the positive side, efforts are being made ity will help policy makers answer important to address some of the data-related problems questions, including the following: listed above. One of such projects is the West a. In terms of labor mobility across space, Africa Survey Harmonization Project, which how much and how fast have people been supports the eight member states of WAEMU, migrating from rural to urban areas or including Benin, Burkina Faso, Côte d’Ivoire, and from one country to another? What are Togo, and conducts harmonized and comparable the implications of any such develop- household surveys using regional standards. The ments on welfare in the host locations? outcomes of this project will help policy makers Is the infrastructure ready and adequate monitor progress in poverty reduction as well as to absorb future inflows of people given improvements in agricultural productivity over current speed? A related study of migra- time and promote regional economic integration tion could also focus on constraints on among member countries. greater and more accelerated agglomera- To deepen our understanding of the economic tions, including secondary cities, and aim geography of the subregion and to hold precise to inform investment projects in urban and relevant policy discussions at country levels, areas. further research is needed in the following areas: b. In terms of labor mobility across sectors, 1. Geographic poverty traps: A crucial question what patterns and trends are observable for any policy designed to tackle poverty is as regards to structural change in each why poor locations stay poor over time, i.e., country? Has there been a transition of why poverty traps persist. The concept of labor out of agriculture with the services poverty traps can be understood as a set of or manufacturing sectors absorbing this self-reinforcing mechanisms whereby a loca- labor? Within each sector, has there been tion starts out poor and remains poor. In other a transition from low-productivity to high- words, current poverty is itself a direct cause productivity jobs? Most importantly, what of poverty in the future (Azariadis and Sta- has been the role of structural transfor- churski, 2005). While the literature proposes mation in poverty reduction? various explanations such as restrictions on 3. Agricultural productivity: Boosting agri- labor mobility (Jalan and Ravallion, 2002) and cultural productivity is a policy priority for limited availability of production technologies alleviating poverty and reducing intraregional that can lead to higher income outcomes in income gaps. Our current analysis focuses on poor areas (Kraay and McKenzie, 2014), it is agroecological characteristics, agglomera- important to study the specific mechanisms tion, market access, and three farm inputs of poverty traps in each of the four countries (spending on fertilizer, land tenure, and land in order to propose meaningful interventions. size). However, in-depth research is needed The outcomes from the West Africa Survey Policy Recommendations and Further Studies 61 to fully understand the determinants of agri- introduced in this report. In particular, such an cultural productivity. Such a study should index should take into account access to all investigate various measures of agricultural transportation modes including coastal ports, productivity (yields and production value) of airports, roads, railways, and waterways as all major crops and explore an extensive set well as access to markets across borders. In of farm inputs such as labor costs, use of addition, the model evaluating the impact of irrigation, and improved seed, etc. Moreover, market accessibility to welfare should aim to such an analysis should be carried out at the address endogeneity that may include better- country level so as to lead to relevant coun- off households choosing to live in locations try-specific policy interventions. For example, with high market accessibility as well as local the classification of agroecological or liveli- governments in poor locations that cannot hood zones can go beyond the four common afford large investments in infrastructure, zones proposed in this report, thus allowing thus suffering from low market access. for detailed policy recommendations on zone- 6. Economic potential across space: This report specific input mixes or R&D. provides stylized facts concerning spatial 4. Climate change and conflict: Political insta- differences in welfare. Another innovative bility and environmental changes can hamper and relevant aspect that helps understand not only welfare in the affected areas but also the spatial landscape of a country would be in the country as a whole. These issues are a study of the potential for rapid economic especially pronounced in West Africa (Marc, development. This study would complement Verjee, and Mogaka, 2015). A complement this report by showing the differences between analysis to this report should overlay maps economic potential and performance across of conflicts and climate changes on maps of the territory. It would be useful for policy mak- poverty, poverty density, and poverty mass. ers to have insights into the geographic distri- Such an analysis should aim to assess the bution of levels of economic potential across roles of political instability and environmental a given country, the relative strengths and changes in spatial inequality in terms of wel- weaknesses of different locations, and the fare. If time series data are available, it would extent to which different locations are fulfill- be critical to understand how these roles have ing their potential. The Economic Potential evolved over time. Moreover, understanding Index (Roberts, 2016) may be one approach how many poor people have remained trapped for taking this analysis forward. This index in poverty in the affected areas over time as captures the extent to which a location pos- well as how poverty density in these loca- sesses five factors that have the potential tions has changed can result in relevant policy to contribute to high levels of productivity. interventions. Related to the labor mobility These five factors (market access, economic study proposed above, the study should focus density, urbanization, skills, and local trans- on a subgroup of migrants, namely refugees; port connectivity) represent key proximate identify who they are in terms of demograph- determinants of local levels of productivity. ics, education, and occupation; and report 7. Regional poverty analysis including Ghana: on the implications of refugee inflows for the With its strategic location in the subregion (i.e., economy. bordered by Côte d’Ivoire, Togo, and Burkina 5. Market accessibility: Market access and its Faso, being located on the same latitude and relationship to poverty are a determinant fac- sharing similar agroecological zones as Côte tor for investment projects in domains such as d’Ivoire, Togo, and Benin), and thanks to its infrastructure and public services. For exam- membership in the Economic Community of ple, an area with a high concentration of poor West African States (ECOWAS), Ghana plays people with low market access may suggest a an important role in the subregional economy. positive return from a road construction proj- Thus, a regional poverty analysis could be ect. For this purpose, we need to improve our enriched if it were to go beyond AFCF2 coun- measures, including the market access index tries and include Ghana. 62 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo References Abro, Zewdu Ayalew, Bamlaki Alamirew Alemu, Sub-Saharan Africa.” Policy Research Working and Munir Hanjra. 2014. “Policies for Agricultural Paper 7729, World Bank, Washington, DC. Productivity Growth and Poverty Reduction in Brakman, Steven, Harry Garretsen, and Charles Rural Ethiopia.” World Development 59: 461–474. van Marrewijk. 2009. The New Introduction to Acemoglu, Daron, Simon Johnson, and James A. Geographical Economics. Cambridge: Cambridge Robinson. 2001. “The Colonial Origins of Com- University Press. parative Development: An Empirical Investigation.” Brinkhoff, Thomas. 2016. citypopulation.de. American Economic Review 91 (5). Buys, Piet, Kenneth M. Chomitz, and Timothy S. Adampoulos, Tasso, and Diego Restuccia. 2017. Thomas. 2005. “Quantifying the Rural-Urban “Geography and Agricultural Productivity: Cross- Gradient in Latin America and the Caribbean.” Country Evidence from Micro Plot-Level Data.” Research Working Paper 3634, World Bank, Working Paper. https://www.economics.utoronto Washington, DC. .ca/diegor/research/GAEZ_paper.pdf. Christiaensen, Luc, Lionel Demery, and Jesper AGRHYMET. 2016. “Côte d’Ivoire: Zones et Kuhl. 2011. “The (Evolving) Role of Agriculture in Descriptions des Moyens d’Existence.” CILSS. Poverty Reduction: An Empirical Perspective.” Akossou, Arcadius, Eloi Attakpa, Noel Fonton, Journal of Development Economics 96: 239–354. Brice Sinsin, and Roel Bosma. 2016. “Spatial and Christiaensen, Luc, and Gabriel Lawin. 2017. Temporal Analysis of Maize Crop Yields in Benin “Maximizing Agriculture’s Contribution to the from 1987 to 2007.” Agricultural and Forest Meteo- Jobs Agenda.” In Toward Better Employment and rology 220 (2016): 177–189. Productive Inclusion: A Jobs Diagnostic for Côte Azariadis, Costas, and John Stachurski. 2005. d’Ivoire, edited by Luc Christiaensen, and Patrick “Poverty Traps.” Chapter 5. In Handbook of Eco- Premand, 68–84. Washington, DC: World Bank. nomic Growth, Vol. 1, Part A, edited by Philippe Combes, Pierre-Philippe, Thierry Mayer, and Aghion and Steven N. Durlauf. Elsevier. Jacques-François Thisse. 2008. Economic Geog- Baldwin, Richard, Rikard Forslid, Philippe Mar- raphy: The Integration of Regions and Nations. Princ- tin, Gianmarco Ottaviano, and Frédéric Robert- eton, NJ: Princeton University Press. Nicoud. 2003. Economic Geography and Public Datt, Gaurav, and Martin Ravallion. 1998. “Farm Policy. Princeton, NJ: Princeton University Press. Productivity and Rural Poverty in India.” Journal of Ballon, Paola, Brian Blankespoor, Yeon Soo Kim, Development Studies 34 (4): 62–85. and Nobuo Yoshida. n.d. “A New Approach for De Brauw, Alan, Valerie Mueller, and Hak Lim Lee. Evaluating the Impact of Transportation Invest- 2014. “The Role of Rural-Urban Migration in the ment on Market Accessibility and Poverty.” Policy Structural Transformation of Sub-Saharan Africa.” Research Working Paper, World Bank, Washing- World Development 63: 33–42. ton, DC. De Janvry, Alain, and Elisabeth Sadoulet. 2010. “Agri- Banerjee, Abhijit, and Esther Duflo. 2007. “The cultural Growth and Poverty Reduction: Additional Economic Lives of the Poor.” Journal of Economic Evidence.” World Bank Research Observer, 25 (1). Perspectives 21 (1): 141–67. Deichmann, Uwe. 1997. Accessibility Indicators in Beegle, Kathleen, Luc Christiaensen, Andrew GIS. New York: United Nations. Dabalen, and Isis Gaddis. 2016. Poverty in a Rising Africa. World Bank: Washington DC. DeLorme. 2015. “Digital Atlas of the Earth 2015.” Berg, Claudia N., Brian Blankespoor, and Harris Diamond, Jared. 1997. Guns, Germs, and Steel. New Selod. 2016. “Roads and Rural Development in York: Norton. References 63 Dixon, Sam, and Julius Holt. 2010. Livelihood Fujita, Masahisa, and Jacques-François Thisse. Zoning and Profiling Report: Burkina Faso. Famine 2002. Economics of Agglomeration. Cambridge: Early Warning Systems Networks Report, USAID, Cambridge University Press. Washington, DC. Funk, Chris, Pete Peterson, Martin Landsfeld, Donaldson, D. 2010. Railroads of the Raj: Estimat- Diego Pedreros, James Verdin, Shraddhanand ing the Impact of Transportation Infrastructure. Shukla, Gregory Husak, James Rowland, Laura American Economic Review. Harrison, Andrew Hoell, and Joel Michaelsen. 2015. “The Climate Hazards Infrared Precipitation with Donaldson, Dave, and Richard Hornbeck. 2016. Stations—A New Environmental Record for Moni- “Railroads and American Economic Growth: A toring Extremes.” Scientific Data 2, article number Market Access Approach.” Working Paper 19213, 150066. National Bureau of Economic Research, Wash- ington, DC. Gallup, John Luke, and Jeffrey D. Sachs. 2001. “The Economic Burden of Malaria” supplement, Easterly, William, and Ross Levine. 2002. “Tropics, American Journal of Tropical Medicine and Hygiene Germs and Crops: How Endowments Influence 64 (1–2). Economic Development.” NBER Working Paper 9106. Gallup, John Luke, Jeffrey D. Sachs, and Andrew D. Mellinger. 1999. “Geography and Economic Devel- Emran, M. Shahe, and Forhad Shilpi. 2012. “The opment.” International Regional Science Review 22 Extent of the Market and Stages of Agricultural (2): 179–232. Specialization.” Canadian Journal of Economics 45 (3): 1125–1153. Goldstein, Markus, Kenneth Houngbedji, Florence Kondylis, Michael O’Sullivan, and Harris Selod. FAO (Food and Agriculture Organization). 2001. 2015. “Formalizing Rural Land Rights in West “Country Pasture/Forage Resource Profiles: Africa: Early Evidence from a Randomized Impact Burkina Faso.” http://www.fao.org/ag/agp/agpc/ Evaluation in Benin.” Policy Research Working doc/Counprof/burkinaFeng.htm. Paper 7435, World Bank, Washington, DC. ———. 2009a. “Country Pasture/Forage Resource Gollin, Douglas, David Lagakos, and Michael E. Profiles: Benin.” http://www.fao.org/ag/agp/agpc/ Waugh. 2014. “The Agricultural Productivity Gap.” doc/counprof/benin/Benin.htm#3climate. Quarterly Journal of Economics 129 (2): 939–993. ———. 2009b. “Country Pasture/Forage Resource Gollin, Douglas, Martina Kirchberger, and David Profiles: Côte d’Ivoire.” http://www.fao.org/ag/ Lagakos. 2016. “Living Standards across Space: agp/agpc/doc/counprof/ivorycoast/ivorycoast Evidence from Sub-Saharan Africa.” Working .htm#climate. Paper, World Bank, Washington, DC. FEWSNET (Famine Early Warning Systems Net- Golub, Stephen. 2012. “Entrepot Trade and Smug- work). 2016. gling in West Africa: Benin, Togo, and Nigeria.” The Foster, Vivien and Cecillia Briceño-Garmendia. World Economy 35 (9): 1139–1161. 2010. Africa’s Infrastructure: A Time for Transfor- Govereh, Jones, and T. S. Jayne. 2003. “Cash Crop- mation. Washington, DC.: Agence Française de ping and Food Crop Productivity: Synergies or Développement and World Bank. Trade-offs?” Agricultural Economics 28 (1): 39–50. Fuglie, Keith, and Nicholas Rada. 2012. “Con- doi:10.1111/j.1574-0862.2003.tb00133.x. straints to Raising Agricultural Productivity in Govereh, Jones, T. S. Jayne, and James Nyoro. Sub-Saharan Africa.” In Productivity Growth in 1999. “Smallholder Commercialization, Inter- Agriculture: An International Perspective, edited by linked Markets and Food Crop Productivity: Cross- K. Fuglie, S. Wang, and V. Eldon Ball. Oxfordshire, country Evidence in Eastern and Southern Africa.” UK: CAB International. http://aec3.aec.msu.edu/fs2/ag_transformation/ Fujita, Masahisa, Paul R. Krugman, and Anthony J. atw_govereh.pdf. Venables. 1999. The Spatial Economy: Cities, Grosh, Margaret, Carlo del Ninno, Emil Tesliuc, Regions, and International Trade. Cambridge, MA: and Azedine Ouerghi. 2008. For Protection and MIT Press. 64 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo Promotion: The Design and Implementation of Effec- Kraay, Aart, and David McKenzie. 2014. “Do Pov- tive Safety Nets. World Bank, Washington, DC. erty Traps Exist? Assessing the Evidence.” Journal of Economic Perspectives 28 (3): 127–148. Hanson, Gordon. 2005. “Market Potential, Increas- ing Returns, and Geographic Concentration.” Jour- Krugman, Paul. 1991. “Increasing Returns and nal of International Economics 67: 1–24. Economic Geography.” Journal of Political Economy 99 (3): 483–99. Harris, Chauncy D. 1954. “The Market as a Fac- tor in the Localization of Industry in the United ———. 1996. The Self-Organizing Economy. Oxford: States.” Annals of the Association of American Geog- Blackwell. raphers 44: 315–348. Krugman, Paul R., and Anthony J. Venables. 1995. HarvestChoice, International Food Policy Research “Globalization and the Inequality of Nations.” The Institute (IFPRI), and University of Minnesota. Quarterly Journal of Economics 110: 857–880. 2016. “CELL5M: A Multidisciplinary Geospatial Lall, Somik V., J. Vernon Henderson, and Anthony Database for Africa South of the Sahara.” Har- J. Venables. 2017. Africa’s Cities: Opening Doors to vard Dataverse, V2. http://dx.doi.org/10.7910/ the World. World Bank, Washington, DC. DVN/G4TBLF. Lall, Somik V., Zmarak Shalizi, and Uwe Deich- Hausmann, Ricardo. 2001. “Prisoners of Geogra- mann. 2004. “Agglomeration Economies and Pro- phy.” Foreign Policy 122, 44–53. ductivity in Indian Industry.” Journal of Development Henderson, Vernon. 2014. “Urbanization and the Economics 73 (2004): 643–73. Geography of Development.” Policy Research Lee, Jonq-Ying, and Mark G. Brown. 1989. “Con- Working Paper 6877, World Bank, Washington, DC. sumer Demand for Food Diversity.” Southern Jour- Hijmans, Robert J., Susan E. Cameron, Juan L. nal of Agricultural Economics 1989: 47–53. Parra, Peter G. Jones, and Andy Jarvis. 2005. “Very Ligon, Ethan, and Elisabeth Sadoulet. 2007. High Resolution Interpolated Climate Surfaces for “Estimating the Effects of Aggregate Agricultural Global Land Areas.” International Journal of Climatol- Growth on the Distribution of Expenditures.” Back- ogy 25: 1965–1978. ground Paper. WPR 2008. Irz, Xavier, Lin Lin, Colin Thirtle, and Steve Wig- Marc, Alexandre, Neelam Verjee, and Stephen gins. 2001. “Agricultural Productivity Growth and Mogaka. 2015. The Challenge of Stability and Poverty Alleviation.” Development Policy Review 19 Security in West Africa. Washington, DC: World (4): 449–466. Bank and Agence Française de Développe- ISciences LLC. 2008. Elevation and Depth v2 2000 ment. https://openknowledge.worldbank.org/ (dataset). http://geoserver.isciences.com:8080/ handle/10986/22033. geonetwork/srv/en/resources.get?id=208&fnam Marshall, Alfred. 1920. Principles of Economics. e=ElevationGridv2.zip&access=private London: MacMillan. Jalan, Jyotsna, and Martin Ravallion. 2002. Mayer, Thierry. 2008. “Market Potential and “Household Income Dynamics in Rural China.” In Development.” Working Paper DP 6798, Center Insurance Against Poverty, edited by Stefan Dercon, for Economic Policy Research, Washington, DC. 108–124. Oxford: Oxford University Press. McMillan, Margaret, and Kenneth Harttgen. 2014. Jedwab, Remi, and Adam Storeygard. 2016. “The “What is Driving the ‘African Growth Miracle?’” Heterogeneous Effects of Transportation Invest- Working Paper 20077, National Bureau of Eco- ments: Evidence from Sub-Saharan Africa.” nomic Research, Washington, DC. mimeo, presented at the GWU/World Bank 3rd Urbanization and Poverty Reduction Research Ministère de l’Environnement et des Ressources Conference (February 1, 2016). Forestières. 2003. “Stratégie de Conservation et d’Utilisation Durables de la Diversité Biologique.” Kottek, Markus, Jurgen Grieser, Christoph Beck, Lomé. Bruno Rudolf, and Franz Rubel. 2006. “World Map of the Köppen-Geiger climate classification ———. 2014. “Cinquième Rapport National sur la updated.” Meteorol. Z., 15, 259–263. Diversité Biologique du Togo 2009–2014.” Lomé. References 65 Minten, Bart, and Christopher Barrett. 2008. Sachs, Jeffrey D., and Pia Malaney. 2002. “The “Agricultural Technology, Productivity, and Pov- Economic and Social Burden of Malaria.” Nature erty in Madagascar.” World Development 36 (5): Insight 415 (6872). 797–822. ———. 2011. “Agricultural Productivity and Poverty Minten, Bart, Lalaina Randrianarison, and Reduction: Linkages and Pathways.” Evans School Johan F. M. Swinnen. 2009. “Global Retail Chains Review 1 (1): 56–74. and Poor Farmers: Evidence from Madagascar.” Smith, Adam. 1776. The Wealth of Nations. New World Development 37 (11): 1728–1741. doi:10.1016/ York: Simon & Brown. j.worlddev.2008.08.024. Strasberg, Paul J., Thomas S. Jayne, Takashi Nunn, Nathan and Diego Puga. 2012. “Ruggedness: Yamano, James K. Nyoro, Daniel David Karanja, The blessing of bad geography in Africa.” Review of and John Strauss. 1999. “Effects of Agricultural Economics and Statistics, 94(1): 20–36. Commercialization on Food Crop Input Use and OECD (Organisation for Economic Co-operation Productivity in Kenya.” MSU International Devel- and Development). 1994. Creating Rural Indica- opment Working Paper 71. tors for Shaping Territorial Policies. Paris: OECD Thériault, Veronique, Alpha Kergna, Abramane Publications. Traore, Bino Teme, and Melinda Smale. 2015. Porter, Michael. 1998. “Clusters and the New Eco- “Review of the Structure and Performance of the nomics of Competition.” Harvard Business Review, Fertilizer Value Chain in Mali.” Food Security Policy November–December, 77–90. Innovation Lab Working Paper Mali-2015-2, Michi- gan State University, East Lansing, MI. Puga, Diego. 1999. “The Rise and Fall of Regional Inequalities.” European Economic Review 43: Uchida, Hirotsugu, and Andrew Nelson. 2010. 303–334. “Agglomeration Index: Toward a New Measure of Urban Concentration.” Working Paper 2010/29, Roberts, Mark. 2016. “Identifying the Economic United Nations University, Tokyo. Potential of Indian Markets.” World Bank Policy Research Working Paper 7623. van de Walle, Dominique. 2013. “Lasting Welfare Effects of Widowhood in Mali.” World Development Rodrik, Dani, Arvind Subramanian, and Francesco 51: 1–19. Trebbi. 2004. “Institutions Rule: The Primacy of Institutions over Geography and Integration in ———. 2015. “Women Left Behind? Poverty and Economic Development.” Journal of Economic Headship in Africa.” Working Paper 7331, World Growth 9 (2): 131–165. Bank, Washington, DC. Rosenthal, Stuart S., and William C. Strange. Venables, Anthony J. 1996. “Equilibrium Locations 2004. “Chapter 49 Evidence on the Nature and of Vertically Linked Industries.” International Eco- Sources of Agglomeration Economies.” In Hand- nomic Review 37: 341–59. book of Regional and Urban Economics, Vol. 4, 2119– Vissoh, Pierre V., Gualbert Gbéhounou, Adam 71. Amsterdam: Elsevier. Ahanchede, Thom W. Kuyper, and Niels G. Röling. Sachs, Jeffrey D. 2000. “Tropical Underdevelop- 2004. “Weeds as Agricultural Constraint to Farm- ment.” Prepared for Economic History Association ers in Benin: Results of a Diagnostic Study.” Wagnin- Annual Meeting. CID Working Paper 57. http:// gen Journal of Life Science 52 (3/4): 305–329. www.cid.harvard.edu/cidwp/057.htm. 66 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo World Bank. 2008. Agriculture for Development. ———. 2017. World Development Indicators. Wash- World Development Report. Washington, DC: ington, DC: World Bank. World Bank. World Food Programme. 2014. “E-voucher for Food ———. 2009. Reshaping Economic Geography. World Security—A Potential for India’s Social Safety Development Report. Washington, DC: World Nets?” World Food Programme. Bank. Yoshida, Nobou, and Uwe Deichmann. 2009. “Mea- ———. 2010. The Poor Half Billion in South Asia— surement of Accessibility and Its Applications.” What is Holding Back Lagging Regions? Washington, Journal of Infrastructure Development 1 (1): 1–16. DC: World Bank. Policy Recommendations and Further Studies 67 Appendix A Agroecological Zones Map A.1  Benin Sources: Data from FAO (2009a), image from Akossou et al. (2016). Map A.2  Burkina Faso Source: FEWSNET (2016). Map A.3  Togo Source: Ministère de l’Environnement et des Ressources Forestières (2014). Agroecological Zones 69 Map A.4  Côte d’Ivoire Source: AGRHYMET (2016). 70 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo Appendix B Market Accessibility Index—Methodology Measuring Access equation (2) and employ the parameters of a = 20 and 30, and b = 2. We then summarize market to Regional Markets access at an administrative level for each country by transforming the market access results from Following the standard approach to calculating both functional forms to an inverse distance access to markets in the literature, the domes- weighted grid and taking the mean of the grid in tic market access for a given location along the the administrative level for each model. road network is a function of the weighted sum of populated places in all other locations discounted by travel time on the road.20 Formally, we define market access in a location i (MAi): Data Pj The administrative boundary data are taken from MAi = ∑ (1) -q j t ij the statistical services of each country or from a � global database in order to match the boundaries where Pj is the population in location j, t ij is with the aggregated census data. The boundary travel time between locations i and j, and q is a files of 37 prefectures in Togo are taken from the trade elasticity parameter. Following Donaldson National Statistical Agency. The 2015 boundary (2010), we use elasticity of trade, q, equal to 3.8 document for 108 departments in Côte d’Ivoire for equation (1) of the classical model. Following is taken from the National Institute for Statis- other regional work with geographically limited tics. The boundary document for 78 communes populated places data (e.g., Lall, Shalizi, and Deich- in Benin is taken from FAO.21 The 2006 bound- mann, 2004; Yoshida and Deichmann, 2009; Bal- ary document for 45 provinces in Burkina Faso lon et al., n.d.), we define another market access is taken from the National Institute for Statistics according to the Negative Exponential model (see and Demography. Deichmann, 1997) in location i (MAi): We used georeferenced city-level estimated popu- − t ij− b lation data from the set compiled by Blankespoor, ( ) MAi = ∑ Pj ⋅ e 2a2 (2) Khan, and Selod (n.d.). The primary source for their � j city population levels data is the census estimates compiled by Brinkhoff (2016) and subsequently where Pj is the population in location j, tij is travel georeferenced in order to add the spatial dimen- time between locations i and j, and a and b are sion. These data provide estimates for 2015 by trade elasticity parameters based on Deichmann using a constant continuous growth rate for cit- (1997). We use the negative exponential model in ies derived from intercensal population data. We impose a minimum population of greater than or 20 Examples from the literature with similar market access include Harris (1954); Hanson (2005); Emran and Shilpi (2012); Jedwab and Storeygard (2015); Blankespoor et al. (2016); Berg, Blankespoor, and 21 We merge the urban and rural boundaries for Djougou commune Selod (2016); and Donaldson and Hornbeck (2016). in Benin in order to harmonize the data with development indicators. Table B.1  Population in City Database for Each Country Country 2015 Population (WDI) Share of Urban Population (%) Benin 10,879,829 0.357 Burkina Faso 18,105,570 0.278 Côte d’Ivoire 22,701,556 0.460 Togo 7,304,578 0.351 equal to 10,000, which yields 206 cities in the near function to force the city locations to coin- four countries under study as locations of regional cide with the nearest road node. Second, we con- markets.22 The sum of population from these loca- struct additional nodes from the intersection of tions captures between 27 and 46 percent of the a 5-kilometer radius from each city and the radial total country-level population from World Bank roads in order to ensure a wider geographic cover- World Development Indicators (WDI) (Table B.1), age of model results. which suggests that the rural population makes up a large share of the total and that the available data do not sufficiently account for local markets. We modify the coordinates slightly so that each Results: Access to Regional populated location corresponds to a node on the Markets road network in order to calculate market access in equation (1) (see above). Both models produce a skewed distribution of market access across the four countries. The The roads data we used are from DeLorme (2015), capital and large cities have primary regional radial which provides a regionally consistent and well- roads, whereas the majority of administrative connected geometry of road segments for net- regions do not have strong connections to these work analysis. The functional road categories in regional markets (proxied by the set of approxi- the road data include primary, secondary, and mately 200 cities). tertiary roads. Derived from estimates in Jedwab and Storeygard (2016), we assume the following Following recent literature using the classical speeds by road categories: 60 km/h for primary model (Jedwab and Storeygard, 2015; Berg, Blank- roads, 40 km/h for secondary roads,23 12 km/h for espoor, and Selod, 2016; Blankespoor et al. 2016), tertiary roads, and 5 km/h as background travel a robustness check with alternative values (e.g., speed in the absence of a road.24 8.2) of the trade elasticity provides similar results to theta 3.8.25 Following the negative exponen- Finally, we make a number of modifications to the tial model, a robustness check provides similar input data. First, we combine the spatial distribu- results for a = 20 and a = 30. When comparing the tion of the cities and road segments by using a two models, Burkina Faso and Côte d›Ivoire have a strong correlation to themselves compared to the results from Benin and Togo.26 22 We exclude five cities that lack intercensal population data on the citypop website. 23 Since DeLorme (2015) does not include the surface type of the road, we use 50km/h as the average of Jedwab and Storeygard’s (2016) estimates for paved road and improved road. 25 The model results with theta 3.8 or 8.2 have a correlation of 0.98 24 We also consulted and made minor modifications based on the or greater by country. road data available from the Africa Infrastructure Country Diagnostic 26 The correlations between the two models are: BFA 0.94, CIV 0.95, Study (Foster and Briceño-Garmendia, 2010). BEN 0.73, and TGO 0.74. 72 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo Appendix C Extra Materials Figure C.1  Poverty Rates by Administrative Unit Bénin cotonou seme−kpodjl porto novo perere bohicon parakou adjohoun save ndali ouidah adjarra abomey−calavi sakete bonou natingou bassila malanville ouesse dassa abomey grand−popo avrankou kouande bante savalou adja−ouere glazoue dangbo come kerou akpro−misserete ketou tchaourou agbangnizoun so−ava kandi lalo cove lokossa houeyogbe aguegue ouake toviklin ifangni toffo djidja karimama tanguieta zogbodome allada athieme bopa za kpota gogounou dogbo−tota banikoara pobe tchoukoutouna sinende bembereke zangnanado ze aplahoue tori−bossito kpomasse nikki djakotome pehonko kalale djougou ouinhi klouekanme materi segbana cobli copargo boukoumbe 0 .2 .4 .6 .8 1 Poverty rates Burkina Faso noumbiel nahouri kadiogo seno oudalan comoe yagha poni boulgou sanmatenga kompienga houet leraba soum gourma ganzourgou oubritenga zoundweogo koulpelogo bougouriba sanguie kenedougou mouhoun namentenga banwa gnagna kouritenga boulkiemde tuy ziro bazega nayala sissili tapoa kossi bale passore yatenga ioba bam loroum kourweogo komandjoari zondoma sourou 0 .2 .4 .6 .8 Poverty rates Togo golfe lacs lome commune ogou cinkasse vo bassar tchaoudjo danyi yoto wawa ave zio moyen−mono agou tone kloto kpele kozah haho assoli sotouboua binah est−mono bas−mono tchamba anie kpendjal amou keran oti dankpen blitta tandjoare akebou doufelgou 0 .2 .4 .6 .8 1 Poverty rates 74 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo Côte d’Ivoire gueyo abidjan tabou soubre gagnoa bangolo blolequin bouafle attiegouakro lakota bettie dabou yakasse−attobrou yamoussoukro doropo san−pedro guiglo dimbokro arrah prikro aboisso transua bouake divo adzope zuenoula jacqueville kounahiri abengourou sikensi buyo bouna niakaramadougou zoukougbeu dabakala bongouanou adiake tiassale meagui daloa daoukro katiola duekoue koun−fao sakassou sandegue korhogo vavoua tai ferkessedougou grand−bassam issia m’bengue taabo bondoukou didievi man dikodougou kouto beoumi dianra agboville seguela toumodi kani guitry sassandra fresco tiebissou kouassi−kouassikro sinfra kouibly alepe tanda mankono kaniasso m’bahiakro bocanda touba agnibilekrou koro facobly tiapoum m’batto akoupe danane sinematiali grand−lahou biankouma samatiguila odienne botro ouangolodougou zouan−hounien boundiali kong djekanou seguelon tehini ouaninou madinani toulepleu nassian minignan tengrela sipilou oume 0 .2 .4 .6 .8 Poverty rates Extra Materials 75 Figure C.2  Poverty Density by Administrative Unit Benin karimama tanguieta bassila ndali perere save kerou segbana gogounou tchaourou kouande bante ouesse tchoukoutouna bembereke malanville sinende savalou pehonko kandi dassa natingou nikki kalale djidja glazoue banikoara materi ketou djougou boukoumbe copargo cobli ouake zangnanado cove zogbodome bonou adjohoun grand−popo sakete ze tori−bossito toffo ouinhi aplahoue adja−ouere athieme lalo kpomasse agbangnizoun bopa ouidah parakou pobe houeyogbe allada za kpota dogbo−tota seme−kpodji lokossa klouekanme aguegue come ifangni abomey so−ava dangbo bohicon djakotome toviklin adjarra abomey−calavi avrankou akpro−misserete porto novo cotonou 0 500 1,000 1,500 2,000 Number of poor people per square kilometer 76 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo Burkina Faso noumbiel kompienga oudalan nahouri yagha comoe seno poni soum gourma komandjoari leraba tapoa kenedougou bougouriba sanmatenga sissili ziro boulgou mouhoun tuy koulpelogo banwa houet kossi namentenga nayala gnagna sanguie loroum bale zoundweogo ganzourgou sourou bazega oubritenga ioba bam yatenga passore kadiogo boulkiemde kourweogo kouritenga zondoma 0 20 40 60 80 Number of poor people per square kilometer Togo bassar ogou sotouboua est−mono tchamba tchaoudjo assoli keran blitta danyi wawa oti agou kpele dankpen ave haho anie amou cinkasse akebou kpendjal yoto moyen−mono lacs zio doufelgou vo binah tandjoare kozah tone kloto bas−mono golfe lome commune 0 500 1,000 1,500 Number of poor people per square kilometer Extra Materials 77 Côte d’Ivoire bouna doropo tabou dabakala niakaramadougou kani kong prikro sandegue seguelon kaniasso tai blolequin odienne madinani minignan gueyo tehini koro koun−fao yakasse−attobrou bettie attiegouakro kounahiri m’bengue touba dianra arrah dikodougou katiola daoukro guiglo nassian ferkessedougou seguela samatiguila dimbokro kouto ouaninou lakota m’bahiakro soubre mankono bondoukou transua aboisso tiebissou sakassou zuenoula kouassi−kouassikro boundiali alepe bouafle zoukougbeu jacqueville didievi san−pedro fresco taabo dabou toumodi tiassale vavoua tanda abengourou gagnoa beoumi biankouma adzope buyo guitry m’batto bocanda divo korhogo bangolo sikensi agboville adiake sassandra bongouanou ouangolodougou sipilou issia akoupe grand−lahou facobly botro duekoue djekanou meagui tengrela danane tiapoum daloa man bouake agnibilekrou toulepleu kouibly yamoussoukro sinfra sinematiali grand−bassam zouan−hounien oume abidjan 0 50 100 150 200 Number of poor people per square kilometer 78 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo Figure C.3  Key Crops across Zones Proportion of individuals planting a crop across zones Extra Materials 79 Table C.1  Proportion of Individuals Growing Crops across Zones (top 5 crops) Côte d’Ivoire Burkina Faso Togo Benin Proportion Proportion Proportion Proportion Zone Crop (%) Crop (%) Crop (%) Crop (%) 1 Yam 67 Mil 76 Maize 89 Cereals 95 1 Cashew 45 Sorghum 70 Sorghum 59 Cotton 68 1 Cassava 18 Cowpea 33 Beans, 53 Tubers 53 cowpea 1 Peanut 15 Peanut 28 Okra 50 Fruits and 47 vegetables 1 Cocoa 14 Sesame 20 Rice 46 Palm oil 5 2 Maize 56 Sorghum 88 Maize 88 Cereals 92 2 Fluvial 49 Cowpea 72 Beans, 45 Tubers 83 rice cowpea 2 Cotton 49 Mil 63 Sorghum 44 Cotton 43 2 Cashew 42 Peanut 60 Yam 43 Fruits and 42 vegetables 2 Peanut 39 Maize 33 Soya 34 Palm oil 2 3 Cocoa 73 Maize 72 Maize 93 Cereals 91 3 Yam 24 Sorghum 71 Beans, 66 Tubers 85 cowpea 3 Rice 23 Peanut 39 Cassava 42 Fruits and 61 (bas vegetables fond) 3 Cassava 21 Mil 34 Yam 39 Palm oil 16 3 Coffee 20 Cowpea 30 Sorghum 22 Cotton 15 4 Cocoa 72 Maize 83 Maize 94 Cereals 91 4 Yam 18 Sorghum 58 Cassava 45 Tubers 76 4 Rice 18 Cotton 38% Palm 26 Fruits and 60 (bas nuts vegetables fond) 4 Cassava 16 Peanut 36% Peanut 23 Palm oil 49 4 Coffee 13 Sesame 33% Beans, 22 Cotton 3 cowpea 80 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo Table C.2  Poverty and Other Indicators, by Zone, with Standard Errors Benin Burkina Faso Cote d/Ivoire Togo Zone 1 Zone 2 Zone 3 Zone 4 Zone 1 Zone 2 Zone 3 Zone 4 Zone 1 Zone 2 Zone 3 Zone 4 Zone 1 Zone 2 Zone 3 Zone 4 Poverty 0.776 0.831 0.659 0.619 0.354 0.463 0.424 0.429 0.302 0.442 0.335 0.212 0.746 0.598 0.603 0.237 rates (0.012) (0.007) (0.008) (0.005) (0.015) (0.007) (0.013) (0.009) (0.009) (0.010) (0.010) (0.006) (0.023) (0.026) (0.021) (0.013) Access to services Cellphone 0.733 0.719 0.790 0.825 0.804 0.863 0.762 0.749 0.664 0.699 0.654 0.814 0.576 0.751 0.659 0.640 (0.015) (0.009) (0.009) (0.005) (0.023) (0.008) (0.020) (0.014) (0.020) (0.016) (0.021) (0.013) (0.033) (0.037) (0.031) (0.036) Electricity 0.136 0.143 0.285 0.403 0.014 0.196 0.052 0.169 0.518 0.468 0.396 0.771 0.123 0.443 0.272 0.733 (0.009) (0.006) (0.008) (0.005) (0.004) (0.006) (0.006) (0.007) (0.009) (0.010) (0.011) (0.006) (0.017) (0.027) (0.019) (0.013) Improved 0.136 0.139 0.284 0.511 0.154 0.573 0.187 0.565 0.631 0.625 0.688 0.882 0.167 0.432 0.248 0.835 toilet (0.009) (0.006) (0.007) (0.005) (0.011) (0.007) (0.010) (0.009) (0.009) (0.010) (0.010) (0.004) (0.019) (0.027) (0.019) (0.011) Piped water 0.214 0.194 0.248 0.411 0.014 0.175 0.016 0.092 0.329 0.113 0.118 0.510 0.065 0.323 0.248 0.399 (0.011) (0.007) (0.007) (0.005) (0.004) (0.006) (0.003) (0.005) (0.009) (0.006) (0.007) (0.007) (0.013) (0.025) (0.019) (0.015) Share of employed working-aged individual in each sector Agriculture 0.716 0.773 0.641 0.386 0.951 0.728 0.921 0.796 0.535 0.565 0.589 0.378 0.491 0.385 0.628 0.097 (0.013) (0.008) (0.008) (0.005) (0.007) (0.007) (0.007) (0.008) (0.010) (0.010) (0.011) (0.007) (0.040) (0.030) (0.030) (0.011) Industry 0.065 0.066 0.118 0.226 0.017 0.069 0.014 0.060 0.122 0.197 0.137 0.191 0.075 0.159 0.133 0.264 (0.007) (0.005) (0.006) (0.004) (0.004) (0.004) (0.003) (0.005) (0.006) (0.008) (0.008) (0.006) (0.021) (0.023) (0.021) (0.016) Service 0.166 0.119 0.149 0.273 0.032 0.202 0.065 0.143 0.338 0.236 0.270 0.424 0.311 0.354 0.188 0.469 (0.011) (0.006) (0.006) (0.005) (0.006) (0.006) (0.007) (0.007) (0.009) (0.009) (0.010) (0.007) (0.037) (0.030) (0.025) (0.018) Other 0.052 0.042 0.093 0.116 0.000 0.000 0.000 0.000 0.005 0.002 0.004 0.007 0.122 0.102 0.050 0.170 sector (0.007) (0.004) (0.005) (0.003) 0.000 0.000 0.000 0.000 (0.001) (0.001) (0.001) (0.001) (0.026) (0.019) (0.014) (0.014) Housing conditions House 0.939 0.952 0.874 0.880 0.608 0.784 0.690 0.586 0.496 0.626 0.571 0.238 ownership (0.008) (0.004) (0.007) (0.005) (0.020) (0.014) (0.021) (0.017) (0.034) (0.042) (0.032) (0.032) Concrete 0.006 0.008 0.010 0.028 0.000 0.003 0.003 0.002 0.011 0.002 0.007 0.007 0.003 0.004 0.001 0.138 roof (0.003) (0.002) (0.002) (0.002) (0.001) (0.001) (0.002) (0.002) (0.004) (0.002) (0.004) (0.003) (0.004) (0.006) (0.002) (0.026) Extra Materials Concrete 0.191 0.395 0.477 0.449 0.024 0.054 0.061 0.116 0.588 0.544 0.392 0.500 0.058 0.186 0.213 0.617 wall (0.014) (0.010) (0.011) (0.007) (0.009) (0.006) (0.011) (0.010) (0.021) (0.017) (0.022) (0.017) (0.016) (0.034) (0.027) (0.037) (continued) 81 82 Table C.2  Continued Benin Burkina Faso Cote d/Ivoire Togo Zone 1 Zone 2 Zone 3 Zone 4 Zone 1 Zone 2 Zone 3 Zone 4 Zone 1 Zone 2 Zone 3 Zone 4 Zone 1 Zone 2 Zone 3 Zone 4 Household demographics Household 7.453 7.432 6.662 5.629 12.508 14.290 10.847 11.423 7.043 8.227 5.501 6.558 8.461 7.421 7.162 6.825 size (0.130) (0.069) (0.070) (0.037) (0.333) (0.196) (0.198) (0.210) (0.138) (0.171) (0.108) (0.102) (0.254) (0.289) (0.260) (0.292) Share of 0.440 0.431 0.487 0.514 0.443 0.494 0.434 0.489 0.540 0.535 0.589 0.613 0.447 0.511 0.485 0.566 working-age (0.005) (0.003) (0.004) (0.002) (0.005) (0.003) (0.005) (0.003) (0.004) (0.005) (0.005) (0.003) (0.009) (0.012) (0.009) (0.007) Dependency 1.780 1.774 1.531 1.488 1.666 1.575 1.850 1.492 1.737 1.505 1.370 1.412 1.630 1.536 1.487 1.416 ratio (0.034) (0.021) (0.022) (0.014) (0.053) (0.021) (0.049) (0.029) (0.053) (0.032) (0.045) (0.035) (0.071) (0.088) (0.059) (0.075) Household head demographics Male 0.934 0.926 0.842 0.802 0.958 0.944 0.936 0.912 0.715 0.915 0.832 0.828 0.888 0.757 0.746 0.676 (0.008) (0.005) (0.008) (0.006) (0.012) (0.006) (0.011) (0.009) (0.019) (0.009) (0.017) (0.013) (0.021) (0.037) (0.028) (0.035) Married 0.922 0.904 0.878 0.881 0.963 0.921 0.938 0.902 0.662 0.937 0.724 0.771 0.865 0.794 0.896 0.744 (0.009) (0.006) (0.007) (0.005) (0.011) (0.007) (0.011) (0.010) (0.020) (0.008) (0.020) (0.014) (0.023) (0.035) (0.020) (0.033) Standard errors in parentheses.  The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo Appendix D Summary of Findings on Agricultural Activities across Countries, by Zone Côte d’Ivoire Each zone is distinguished by the cash crops produced: cashew in Zone 1, cotton in Zone 2, and cacao in Zones 3 and 4. Zone 1 is widely disadvantaged, with lower yields, less input use, and lower revenues from sales. Zone 1 Cashew is produced in this zone. This is a largely disadvantaged zone. Maize and yam yields are lowest in this zone, as are input and land use (with less spending on fertilizer, less use of pesticides and irrigation, and smaller farms). The value of sales in this zone is also the lowest. Zone 2 Cotton is produced in this zone. This is a zone with the highest input use and largest farm sizes. Alongside Zone 4, this zone has the highest revenues from crop sales. Zones 3 & 4 Cacao is produced in both of these zones, and rice production is also prevalent. Both zones have very high maize yields. Zone 4 has relatively higher revenues from crop sales than Zone 3. Compared to Zones 1 and 2, there is stronger securitization of land in Zones 3 and 4. Burkina Faso Sorghum is prevalent across all zones, and cotton is prevalent in Zone 4, the most advantaged zone. Zone 1 is widely disadvantaged, but has better maize yields than Zones 2 and 3 and better securitization of land than all other zones. Zone 1 This zone has the lowest input use (less spending on fertilizer and less use of pesticides), mainly subsistence farmers (only 30 percent sell some of their crop), and low revenues from sales. However, maize yields are higher than in Zones 2 and 3, and land securitization is stronger than in all other zones. Zones 2 & 3 Maize yields are lowest in these zones, but input use and output sales are higher compared to Zone 1. There is some production of cotton in Zone 3. Zone 4 Cotton is widely produced here. This zone somewhat mirrors Zone 2 in Côte d’Ivoire. This zone has the highest maize yields and input use, the highest proportion of farmers selling some of their crop, and the highest revenues from crop sales. However, securitization of land is the weakest. Togo There is less variation across regions in the dominant crops, with maize, root crops (yam and cassava), and cowpeas being common throughout the country. Pesticide use is low throughout the country. However, Zone 1 has better prospects for agriculture, whereas Zone 2 seems to be disadvantaged. Zone 1 This zone has rice and soya production, and has the largest value of sales. It also has higher spending on fertilizer (comparable to spending on fertilizer in Zone 2). (continued) Zones 2 & 4 These zones have fewer individuals working in agriculture or selling some of their produce. These zones also have lower sales value of produce (but better securitization of land). Zone 2 has higher spending on fertilizer than Zone 4. This zone also has some soya production. Zone 3 This zone has the lowest maize yields as well as low spending on fertilizer. Despite this, the value of sales is higher compared to Zones 2 and 4. Benin The zones are divided across cash crops, with cotton in Zones 1 and 2, palm oil in Zone 4, and to a lesser extent some cotton and palm oil production in Zone 3. In terms of yields, Zones 1 and 2 seem to be more advantaged, whereas Zone 4 less advantaged in terms of yields. Zone 1 Cotton is largely produced in this zone, and with higher yields. Irrigation is slightly higher in Zone 1, but is low throughout the country. Zone 2 Cotton is widely produced in this zone, but with lower yields compared to Zone 1. However, maize yields are highest in this zone. Spending on fertilizer is also highest in this zone. Zone 3 This zone has little cash crop production, but it has some palm oil and cotton production. Zone 4 This zone is prevalent in palm oil production. Maize and cotton yields are the lowest in this zone, as is fertilizer spending. Securitization of land is better in this zone as well as in Zone 3 compared to Zones 1 and 2. 84 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo Appendix E Agricultural Data—Notes on Model Construction For Côte d’Ivoire, there are two relevant parts • Burkina Faso: Conversion into kilograms for to the survey (particularly how the database is nonstandard output units was not available organized): a land section (harvested plots) and a in the survey. Thus, the conversion factor crops section (crops planted on those harvested for a given nonstandard unit was calculated plots). Although these two sections should in the- as the ratio of the average (regional level) ory represent the same number of households, in sales price for a kilogram of maize and the practice they do not. The land section coincides average (regional level) sales price for a given with 6,849 households, while the crops section nonstandard unit of maize. Thus, for Burkina coincides with only 3,402 households. Additional Faso, region-crop-specific conversion factors caution is recommended when interpreting data were used when converting nonstandard units that use the crops section for Côte d’Ivoire (such into kilograms. The same method was used as yields, sales, and input use, but not irrigation, when calculating cotton yields for Burkina which comes from the land section). Faso. • Togo: Conversion into kilograms for nonstandard output units was reported Maize and Cash Crop Yields by respondents in the survey. For each nonstandard unit, the conversion factor was For each country, yield is a measure of average calculated as the regional median of the kilograms per hectare of a crop produced in a given reported conversion factors. Thus, region-crop- agroecological zone; however, due to data con- specific conversion factors were used when straints, the yield measure is calculated slightly converting nonstandard units into kilograms. differently for each country. The conversion factors were self reported as • Côte d’Ivoire: Conversion into kilograms for opposed to assumed from sales and price data. Due to data constraints, assumptions had to nonstandard output units was not available be made on how much land each household in the survey. Thus, the conversion factor allocated to maize (or any given crop). Since the for a given nonstandard unit was calculated survey did not elicit this information directly, as the ratio of the average (national level) we made the assumption that for any given sales price for a kilogram of maize and the plot with more than one crop, the main crop average (national level) sales price for a given is assigned 70 percent of the plot area and nonstandard unit of maize. We chose to the secondary crop is assigned 30 percent calculate a national level conversion factor (as of the plot area. The survey does not contain opposed to district level) due to constraints information on whether a third crop is present on the number of observations for sales of in any given plot. nonstandard units at a subnational level. However, district-level conversion factors • Benin: Yields were not taken from the were used for cacao yields in Côte d’Ivoire. household survey but from data from the Table E.1  Comparison of Yield Data from Ministry of Agriculture and PFR Household Survey Ministry of Agriculture Data (2015) PFR 2011 Data (smallholders) Zone 1 1,280 kg/ha 1,337 kg/ha Zone 2 1,582 kg/ha 1,162 kg/ha Zone 3 1,272 kg/ha 925 kg/ha Zone 4 1,075 kg/ha 1,205 kg/ha Ministry of Agriculture on commune-level from a nonrepresentative household survey estimates of production and land used from 2011 conducted to study the effects of for various crops. The key difference in land titling. comparability is that the Côte d’Ivoire, Overall, the yields of smallholder farmers in 2011 Burkina Faso, and Togo databases may are close to those at the commune level in 2015, reflect smallholder farmers as the data come though patterns may be different across zones. from the household survey, while the Benin Based on 2015 data from the Ministry of Agri- data may reflect larger farms. Caution is culture, Zone 2 had the highest yields and Zone 4 recommended when making comparison of the lowest. Based on the 2011 PFR, Zone 1 had the maize and cotton yields between Benin and highest yields and Zone 3 the lowest. However, the three other countries in the subregion. we cannot draw much from this analysis as the To glean information from Benin on datasets are different in key aspects, including smallholder farmers, we used the PFR 2011 representativeness and time of survey. database, which consists of plot-level data 86 The Geography of Welfare in Benin, Burkina Faso, Côte d’Ivoire, and Togo